Higher Education Administrator Perceptions and Experiences in Predicting and Tracking Retention

Chapter 1: Introduction to the Study

The topic addressed in this study is the application of student data by

administrators of 4-year colleges in tracking and predicting retention. Student data are

derived from student behavior characteristics and attributes that are tracked during their

academic pursuits in higher education. There is a limited number of studies about higher

education application of student data used for tracking and predicting retention. This gap

in the literature affects the information that is needed to address the gap in practice about

processes required to scaffold tracking and predict retention in higher education. A strong

sense of urgency is needed to reaffirm administrators of higher education and their

commitment to using chosen student data to make better strategic decisions, especially as

it relates to student retention. Brown (2019) said that educational leaders should be

advancing using predictive analytics to channel new tools for the success of institutions

and student retention at their chosen university. This study was conducted to address a

gap in literature by informing practices of administrators at 4-year colleges about

processes needed for efficacious student data applications to track and predict retention in

higher education.

This study will contribute to positive social change by increasing higher education

administrators’ understanding of best practices in applying student data targeted to predict

and track retention, ultimately influencing student success and institutional revenue. This

can prove to be valuable and lead to institutional sustainability. An efficacious process for

predicting and tracking retention is an asset in higher education as it allows innovative

recruitment and increases overall efficiency and cost containment when students are

retained. Neelakantan (2019) said the understanding in higher education of student data

for variable identification used for institutional advancement initiative fuels well-

organized processes that allow staff to work smart for increased time management. This

can underpin confidence in the adoption of better decision-making processes as it relates

to student data. Use of policies and procedures that enable data managers to make

informed decisions based on evidence rather than intuition as this aligns with best

practices shown in suggested data analysis processes (Sivarajah, et al., 2017; Schneider &

Preckel, 2017) that are now expected by higher education accrediting bodies, such as The

Higher Learning Commission.

I explore this topic in this chapter. I begin with the study’s background with a

brief description of the focus and scope regarding student data application for retention

tracking in higher education. The problem, purpose, and research question are presented

along with the conceptual framework. The nature of this study and significance of other

research is addressed along with a brief discussion of the study’s assumptions, scope,

and delimitations, followed by a summary of this chapter.

Background

This study involves perceptions and experiences of administrators of 4-year U.S.

colleges regarding identifying student data for tracking and predicting retention. This

study’s scope is appropriate to interpret experiences and perceptions of administrators of

4-year colleges regarding their understanding of the phenomenon that will inform best

practices in terms of the application of student data and retention. The scope of data

collection was not limited by geographical location since these interviews were

conducted in a virtual format as well as face-to-face. Data were collected from 12

administrators from U.S. 4-year colleges who were recruited using criteria-based

sampling.

This study can bridge the gap in literature regarding the application of student

data to track and predict student retention in higher education. Wilderotter (2020) said the

lack of faculty and administrator understanding of one’s student population and their

attributes can prevent selection of variables that measure retention in a meaningful way.

Wilderotter also said critical planning for retention in higher education involves

understanding student enrollment trends and what influences those trends for student

decision-making when choosing a college. Additional research is needed to help higher

education administrators identify predictive student data for retention.

The current study is needed because it will benefit higher educational institutions

and the data administrators that plan for student retention tracking initiatives.

Neelakantan (2019) said returns on investment for intentional planning for student

retention variables outweighs costs. Avella et al. (2016) said that many higher education

administrators are currently unable to communicate the linkage between student data,

identifying variables for tracking student retention. Huda et al. (2018) said that higher

education administrator knowledge of data mining processes identified for retention

efforts drive best practices for tracking and predicting college student retention and

institutional stability through increased funding. This study is needed to further address

gaps in research that influence the gap in practice that is needed for administrators in

higher education to identify and apply student data for higher education internal research

of retention initiatives in 4-year U.S. colleges.

Problem Statement

The problem guiding this study is that many administrators of 4-year colleges do

not have efficacious application practices for student data in terms of tracking and

predicting student retention. Wang (2017) said that despite abundant data availability,

higher education administrators have a limited understanding of student behavior

characteristics that are needed for analyzing and selecting appropriate variables for

overall institutional research. According to Wang, this has a negative effect on data

management practices and student success outcomes, in that the limited understanding of

the data process itself is a hindrance to reliable outcomes.

This study builds upon previous research findings about processes and policies for

identifying student data that can be applied as variables for data analysis as well as using

analytics for tracking and predicting retention in higher education. Sass et al. (2018) said

data mining practices are typically siloed and fluid within institutions and are

characterized by a lack of planning as it relates to internal retention efforts. Huda et al.

(2018) said higher education administrators continue to struggle without guidance on data

management practices for selecting variables from raw data to track and predict student

retention. Further, a lack of knowledge exists among many data administrators in higher

education who lack appropriate data mining policies and processes to drive student

support practices. This lack of data knowledge and practice prevents the successful

function of innovative business intelligence in higher education in terms of retention for

revenue stability (Elhassan et al., 2018).

Purpose of the Study

The purpose of this basic qualitative study is to explore perceptions and

experiences of administrators of 4-year colleges in terms of their application of student

data for tracking and predicting retention. The basic qualitative methodology was used in

this study to generate an increased understanding of this phenomenon. Increased

understanding is needed for higher education administrators for the process of selection

of student data and how they are applied to tracking and predicting retention. This

understanding in higher education for analyzing and selecting data is important because

student attributes that feed into data selection processes may change depending upon

trends influencing higher education (Arensdorf & Naylor-Tincknell, 2016). Matsebula

and Mnkandla (2017) said changing trends can further affect retention tracking because

of the needed reevaluation of targeted student data.

Research Question

The research question that guided this study is:

RQ1: What are the perceptions and experiences of higher education

administrators of 4-year colleges in terms of the application of student data targeted for

tracking and predicting retention?

Conceptual Framework

The process of understanding and interpreting raw data is not reduced to a

predetermined method, but it occurs during the interpretive process itself (Mills et al.,

2010). The primary function of the conceptual framework is to provide a clear and

concise mechanism for interpretation and understanding in a qualitative study (Barbour,

2014). These two models were specifically chosen to underpin this study because they

provide the basic knowledge and defined terms that make up data components of any

retention initiative. Part of gauging the higher education perceptions and experiences in

this study will rely upon their basic knowledge of data mining terms. The conceptual

framework models of Attarran et al. (2018) and Gandomi and Haider (2015) constitute

the data mining and analysis structure needed to inform these perceptions and

experiences. A more detailed analysis of this will be provided in Chapter 2.

The conceptual framework for this study is informed by a model of analytics by

Attaran et al. and the three V construct by Gandomi and Haider. Given the strong

knowledge base needed from participants regarding data mining protocol and language,

these are appropriate conceptual framework models for this study. A more detailed

analysis of this dual lens conceptual framework will be provided in Chapter 2.

The Attaran et al. model takes a three-pronged approach to analytics: descriptive,

predictive, and prescriptive. These three approaches involve decision-making

perspectives seen most frequently in higher education. First, descriptive analytics allows

for a rearview mirror approach to describe data variables chosen that will assist in

choosing a path forward given what has or has not been successful results in the past.

This means that previous data outcomes can be reviewed for historical perspectives and

provide a course forward given past behavior collected from data and whether this

behavior is still wanted or not (Attaran et al., 2018). The second prong of this

threepronged approach is predictive analytics. This particular perspective considers

historical data as well as current data mining results. It is useful in identifying potential

risks as well as opportunities for growth and development with appropriately identified

variables that show patterns of data that serve to predict future behavior of the group

being studied and analyzed with targeted data points. Finally, the prescriptive analytics

tool of this model allows decision-makers a simulation process to optimize an action

being considered in policy or practice. This is a valuable approach and goes beyond

descriptive and predictive analytics by helping decision-makers decide what happens next

and why this course may be best given the data patterns from the group being studied or

analyzed. Kivunja and Kuyini (2017) said that any study based upon basic data mining

skills for the understanding of analysis results and its application in higher education

should have a data mining model as the basis of policy and practice, such as Attaran, et al.

and Gandomi and Haider, that provide the dual lens of the conceptual framework of this

study.

The second model supporting this study is the three-V construct by Gandomi and

Haider. The three Vs stand for: velocity, variety, and volume. Velocity refers to the speed

at which data is constantly uploading into a data system. Variety is diversity of data such

as gender, race, socio-economic status, GPA, and credit ration score (Gandomi & Haider,

2015, p. 138). Volume refers to amount of data, given the size of the institution and

number of goals or targets being measured. The three-V model targets raw and

unstructured data that compose 95% of data analytics in higher education (Mah, 2016), as

well as types of data that higher education administrators and stakeholders consider

during their decision-making processes (Chaurasia & Rosin, 2017). Baer and Norris

(2016) said higher education administrators cannot begin to understand data analytics and

its application without understanding the three V construct.

Exploring perceptions and experiences of administrators of 4-year colleges

regarding their application of student data for tracking and predicting retention requires

basic knowledge of data mining principles such as those supporting this study in the

conceptual framework. Analysis and interpretation are essential to all qualitative inquiries

(Silverman, 2016). During analysis, certain characteristics became more important than

others and meaningful data from participants were driven by the RQ during this analysis.

Further, data analysis was guided by the conceptual framework. It would not be possible

to answer the RQ without these models informing the basic participant knowledge

necessary to scaffold post analysis .

The conceptual framework guides research components of the study. With

understanding of the two models, I created semistructured interview questions. Use of

priori codes is frequently referred to as a deductive or beforehand form of analysis, while

building codes during analysis is inductive (Castillo-Montoya, 2016). I used priori codes

as part of the data analysis procedures. More detail regarding the chosen conceptual

frameworks and their alignment with this study is in Chapter 2.

Nature of the Study

I used a basic qualitative design to explore the perceptions and experiences of

administrators in 4-year colleges about identifying and applying student variables for

tracking and predicting retention. Basic qualitative research is used to examine natural

circumstances in which individuals’ function, as the objective is to provide a practical

understanding of real-world problems (Korstjen & Moser, 2018). Basic qualitative design

is an appropriate choice for this study because this design is used to advance knowledge

and investigate an academic phenomenon (Given, 2008). I investigated the processes in

place for applying student variables in retention, and the perceptions and experiences of

administrators using these practices. Another rationale for choosing the basic qualitative

design is to address contextual experiences of interest. Patton (2015) said the qualitative

inquiry is the chosen method for analyzing people’s perceptions and experiences within

the context to be understood. Accordingly, I analyzed perceptions and experiences of

college administrators.

This design was chosen as opposed to the case study. This is because in the case

study design, a researcher is focusing on the concrete, contextual, in-depth knowledge

about a specific real-world issue that allows a researcher explore the key characteristics,

meanings, and implications of the case (Lichtman, 2020). There is not enough literature

on the topic of this study to inform practice in regards to answering the RQ. Because of

this reason, the researcher found it necessary to have the open design approach that the

basic qualitative study design provides.

The basic qualitative study, sometime called generic, general, or interpretive

qualitative designs typically derives from practical issues in the social sciences which

provide the context for qualitative semistructured interviews. Data were collected via

semistructured interviews by phone or virtual format from administrators at 4-year U.S.

colleges who have responsibility for the identification of student data for retention

efforts. Number of participants depended upon what was required to reach data

saturation. Saturation is achieved when the researcher begins to hear the same comments

repeatedly from participants, and no further interviewing is necessary (Saunders et al.,

2017).

During data analysis, validity involves obtaining rigor through using techniques of

verification such as inductive and deductive coding (Spiers et al., 2018. This approach

has two advantages. First, by using a priori codes or deductive coding, I can ask

participants questions that relate to their own thoughts and experiences that can help

establish rapport as well as gather data. Codes were selected and categorized using NVivo

12 software to ultimately deduct data down to the three major themes. Further discussion

of the basic qualitative study design appears in Chapter 3.

Definitions

Terms that need clarification are discussed briefly here.

Some terms that may need clarification will be discussed briefly here.

Data Mining: This is a general term used frequently when discussing the process

for identifying variables for research. Specifically defined as the practice of analyzing

large databases in order to generate new information (Slater et al., 2017).

Retention: Because this is a common termed heard in both business and education

it is important to have the correct working definition. For this study, student retention in

higher education is defined as the identified student attributes applied as variables in

tracking retention and success of students in higher education from term to term until

completion (Alsharari & Alshurideh, 2020).

Student Data: Mentioned throughout this study as part of the key concepts are

defined as those student data or characteristics deemed statistically significant for use as

variables in tracking a data target (Elegendy & Elragal, 2016).

I assumed that participants shared honest and accurate accounts of their

perceptions and experiences when applying student data for tracking and predicting

retention. It was further assumed that during the participant selection process, participants

accurately portrayed their job titles and responsibilities. This is necessary for internal and

external validity of the study. This is also important in terms of answering the RQ and

relating this knowledge through semistructured interview questions. Magaldi and Berler

(2018) said semistructured interviews are only as strong as identified participants’

knowledge base related to the focal topic. These assumptions were necessary for this

study to build the research question and semi-structured inquiries for participant

interviews.

Scope and Delimitations

The scope of this study includes administrators of 4-year, U.S. colleges who are

responsible for student variable applications for tracking and predicting student retention.

This study was delimited to 4-year colleges to be specific to higher education

administrators without being too narrow (Lichtman, 2017; Patton, 2015). This study was

also delimited to only U.S. colleges. Data collection only within the United States can

minimize efforts for saturation during the COVID-19 crisis that may influence the

participation recruitment process (Bradley et al., 2020). Finlay (2013) stated that in

qualitative study, careful thought for participant selection must be done with the essential

belief that you look for variety in people to describe, explore, or explain phenomena in

real-world contexts. Merriam and Tisdell (2015), and Finlay (2013), stated that

researchers and readers can then make connections from the data outcomes. When these

reflections are applied to qualitative practices, this is called transferability (Barbour,

2014; Finlay, 2013; Given, 2008; Lichtman, 2017; Merriam & Tisdell, 2015).

Limitations

Bias may reduce credibility, and researchers must control for bias during tool

development, sampling, and data interpretation. Researchers should also be aware of

personal thoughts, beliefs, values, and opinions so they do not negatively impact their

study (Creswell & Poth, 2019). I am a retention and completion director in higher

education, and knowledge and responsibilities I have that relate to the focus of this study

can cause bias. I minimized my bias by being cognizant of any influence my background

may have on the study. Potential weaknesses in the study can influence trustworthiness of

the final analysis if not kept in check. I reflected on and attempted to prevent biased

activity. I avoided snowball sampling, which could have resulted in too many likeminded

participants. I also used a reflexivity journal and restricted myself to the research method

process detailed in Chapter 3.

Significance

Findings of this study involve a gap in the literature that is needed to inform

practice related to higher education administrators’ application of student data in

predictive analytics for tracking retention. Hadwater et al. (2019) said that the future

foundation of student retention in higher education will involve policies that informs

analytics practice to save time, money, and human capital. The findings of this study

provide a deeper understanding of processes for identifying student data in predicting and

tracking retention. This, in turn, will benefit higher education administrator practice by

helping to close the gap in literature (Hadwater et al., 2019). According to Rubel and

Jones (2017), to achieve best practices for retention analytics, insight into student

attributes and behavioral characteristics must be applied during the identification of

student data.

Increased understanding needed for higher education administrators on

appropriate student data in predicting retention initiatives can positively influence social

change. The intended audiences for this study are higher education decision makers and

internal research teams that guide analytics interpretation. This can be done both in higher

education and on an individual basis by administrators. Predictive and preventive

analytics mining can bolster retention and completion (Chaurasia et al., 2018). Also,

improved analytics can promote more efficient use of human capital while increasing

revenue (Elhassan Ali & Klett, 2018). This can influence higher education practices to

yield improved institutional stability and inform efforts to reduce siloed data ownership

pools that impede progress (Nimmagadda & Rudra, 2017).

Summary

This chapter included an overview of this study of administrators of 4-year U.S.

colleges and their application of identified student data in predicting and tracking

retention. Further, administrators’ experiences and perceptions of student data identified

as targets in predicting retention through semi-structured interview questions are

grounded in the conceptual framework models of Attaran, et al. (2018) model of analytics

and Gandomi and Haider (2015) three-V construct. Limitations were discussed that

highlighted participants’ lack of knowledge and experience that may influence

recruitment and data outcomes as well as researcher limitations given my current job title

and responsibilities involving institutional retention. Participants were 10 administrators

from eight 4-year U.S colleges . I focused on describing social change implications

relating to improved understanding of higher education administrators with data tracking

responsibilities in predicting retention, which can positively influence student success and

institutional stability by increasing revenue. Chapter 2 contains information regarding the

conceptual framework along with a review of current literature.

Chapter 2: Literature Review

The problem guiding this study is that many administrators in 4-year colleges do

not have efficacious application practices in place for student data in terms of tracking

and predicting student retention. Therefore, the purpose of this basic qualitative study is

to explore perceptions and experiences of 4-year college administrators regarding their

student data application processes for tracking and predicting retention. Problems with

tracking retention arise in higher education without defined processes for applying

student characteristics that are specific to data tracked for students who persist in school

(Elhassan & Klett, 2018; Mahroeian et al., 2017).. Important components of successful

tracking and predicting of retention initiatives in higher education include attitudes and

perceptions of administrators who have the responsibility for pertinent use in defining and

applying student data (Alsharari & Alshurideh, 2020)

In this chapter, research relevant to this study is reviewed. Literature search

strategies and databases used for this research are discussed. I used Stark and Stotler’s

model of analytics and Gandomi and Haider’s three-V construct to address key concepts

in this study. Key concepts addressed through this literature review include defining

retention in higher education, applying and identifying student data for institutional

research, tracking and predicting retention in higher education, and barriers to data

mining. These are followed by chapter conclusions and a summary.

Literature Search Strategy

Databases used during the research process for finding literature were: SAGE

Journals, ERIC, , EBSCOHost, PROQUEST, PsycINFO, Academic Search Premier, and

ProQuest Central. Walden University’s Thoreau Library portal and Google Scholar were

also used to identify articles not obtained through databases. Key words used were:

analytics in higher education, higher education administration decision-making tools,

higher education administration perceptions of student data applied in retention, tracking

and predicting retention in higher education, Gandomi and Haider three-V construct in

higher education, Attaran, Stark, and Stotler analytics model in higher education, student

attributes identified as variables in tracking or predicting retention, qualitative study in

higher education retention, student data in retention, how student data application

influences student retention tracking, higher education student retention, retention

analytics in higher education, and students who retain in higher education. These key

terms were used independently and in combination, and led to four major themes:

defining retention in higher education, applying and identifying student data for

institutional research, tracking and predicting retention in higher education, and barriers

to data mining. There are 60 total sources in the literature review, 41 (68%) of which

were published between 2017 and 2021 (see Table 1).

Table 1

Sources in the Literature Review

Source Type

Last 5 Years

Older than 5 Years

Total

Peer Reviewed

Seminal

Total Sources

Conceptual Framework

The conceptual frameworks for this study were the model of analytics and three V

construct. Both have origins in general data mining best practices (Baker & Siemens,

2014). Baker and Siemens (2014) said accuracy of education data mining (EDM) metrics

is important, as they can determine the relevance of educational experiences to students.

Heiner et al. (2007) said the model of analytics establishes processes and tools needed in

any data mining effort to operationalize decision-making in order to synchronize data for

planned improvement in policy and practice.

Model of Analytics

The model of analytics involves a three-pronged approach to analytics:

prescriptive, predictive, and descriptive. These address the decision-making perspectives

seen most frequently in business and education (Williamson, 2017). Attaran et al. (2018)

said once big data are broken down, these smaller pieces of prescriptive, predictive, and

descriptive data allow decision-makers to see patterns and trends from past data

outcomes.

Descriptive analytics allow for a knowledge-based approach to deciding future

policies and procedures (Attaran et al., 2018). Predictive analytics involves historical data

as well as current data mining results. It is useful in identifying potential risks and

opportunities for growth and development. The prescriptive analytics approach allows

data simulations to optimize an action being considered. Attaran, et al. stated that this is

the most valuable approach and goes beyond descriptive and predictive analytics by

allowing decision-makers to decide what happens next.

Figure 1

Record-Shape Visual

Three V Construct

. The three Vs of the three V construct are velocity, variety, and volume (Gandomi

& Haider, 2015). The speed (velocity) at which student data are available will also

include a assortment of data points and diversity of experiences (variety) with the actual

number, or how much, of analyzed and chosen variables (volume) there were in the final

analysis. Williamson (2017) said the political economy is changing in higher education,

and data science has migrated from the commercial sector into academics. This shift

requires a basic knowledge of how data are defined. Torrecilla and Romo (2018) stated

that one of the models necessary to understand the basics of any data tracking or analysis

is the Gandomi and Haider the V construct.

Summary

The three V construct provides this study with an underpinning for deciphering

the data outcomes as they relate to the perceptions and experiences of administrators in

4year colleges of applying student data for tracking and predicting retention. This is done

by incorporating the three V construct factors of volume, velocity, and variety into the

semistructured interview questions. This construct will assist in data outcome

interpretations in identifying participant experience or perception of data or student

variable identification that relates to its volume, velocity, or variety. Heiner et al. (2007)

stated that Attaran et al.’s model establishes an underpinning for the processes and tools

needed in any data mining effort to operationalize decision making in order to

synchronize data for planned improvement.

Literature Review Related to Key Concepts

The focus of this study will include the application of predictive approaches that

can drive decision making in retention tracking initiatives in higher education and the

efficacious identification of student data in this process. The first subsection of this

review of current literature will provide a review of literature related to the tracking and

predicting of retention in higher education. The ensuing section will review how student

data are identified and used in retention. Next, I discuss literature on the importance of

inclusion in student retention. The main themes that emerged include: defining retention

in higher education, applying and identifying student attributes for institutional research,

tracking and predicting retention in higher education, barriers to data mining.

Defining Retention in Higher Education

Retention in higher education and how certain populations are defined within the

data mining process can be generalized or college specific (Dewberry & Jackson, 2018).

What is noted as the current generalized definition of retention in higher education is the

fall-to-fall comparisons as reported by the National Center for Education Statistics

(2020). This national data clearing house, which also provides the Integrated

Postsecondary Education Data System (IPEDS), is the agency considered as the expert in

the national benchmarks for all higher education (Dewberry & Jackson, 2018; Sass et al.,

2018; Swafford, 2017). Other terms that are associated with retention and have been used

interchangeably are completion, persistence, and student success (Manyanga et al., 2017).

Manyanga et al. (2017) continued and said that many institutions have defined and

delineated these terms differently and a general agreement to the term retention is agreed

upon in research. However, Manyanga et al. cautioned, generalizations about retention

can be misleading due to the uniqueness of each institution, academically, culturally, and

otherwise and should be considered carefully in research efforts.

Prior to the 1970s, student retention was seen as the reflection of individual

attributes, skills, and motivation (Tinto, 2006). Tinto (2006) further stated that students

who did not retain or persist were thought to be less able, less motivated, and less willing

to foresee the fruits of this college labor. In short, students failed, not institutions. That

view, and how institutions define student retention, shifted to consider the role of the

environment, the institution, and level of connectedness a student internalizes for the

decision to stay or leave (Olaya et al., 2020). This student retention definition continued

evolving to what it is today and now has the student variable and data mining descriptors

as part of this explanation (Garcia-Ros et al., 2019).

Nadasen and List (2017) stated that the difficulty is not of the definition of

retention in higher education, it is the lack of information available to track or predict the

retention of students, with a process for data mining and policy to inform best practices.

This is evidenced by Elnozahy et al. (2019) who stated that for new students to retain, the

supportive internal research needs to be accessible to inform practice on student variable

identification for tracking of those existing and retained students fall to fall. This is

discussed more specifically by Mahroeian et al. (2017), who defined and categorized the

perceptions of its participants toward the use of analytics and data mining in New

Zealand higher education. They found public and private 4-year upper management staff

had knowledge of data in three main categories: structural, functional, and

structuralfunctional. Mahroeian et al. further defined these categories via the participants’

perceptions. Those who perceived analytics in terms of the structural aspects leaned

toward quantitative elements such as statistics, digits, visualization, and metrics to inform

decisions. Mahroeian et al., suggested further study to gauge the foundational perceptions

and experiences in higher education related to data mining, use of analytics, and applying

variables. Just as they sought to deliver a better understanding of current perceptions and

values of analytics in higher education within the New Zealand by defining three

functional aspects of data mining, I also seek to do the same regarding efficacious student

variable application in tracking and predicting retention.

Applying and Identifying Student Attributes for Institutional Research

Understanding the student experience from academic, social, and functional

aspects of institutional connectedness can only be recognized and measured if all student

attributes are accounted for in internal research (Stage, 2000). Braxton (2019) suggested

that student data could not be identified for research in retention until specific areas for

future research were addressed. Braxton stated that these areas should be considered for

data mining and defining student attributes: (a) continued study of sociodemographic

characteristics of students, (b) the role of organizational behavior, (c) student

environments nested within different institutional settings, and (d) the effects of student

sub climates (p. 132). However, one such attribute being considered from a psychology

and emotional health point of view, student external commitments, is traditionally not

thought of as a measurement assigned in tracking retention in higher education (Tight,

2020). Tight (2020) continued and stated that institutions continuing to focus on the

obvious attributes such as race, socioeconomic status, SAT, and gender, are doing a

disservice to themselves and students by not attributing external commitment attributes to

retention research tracking.

A trending practice in higher education has been program review (Conrad &

Wilson, 2020). Conrad and Wilson (2020) suggested that institutions consider program

review as an opportunity to assess student retention at program and course levels. In

doing so, Conrad and Wilson continued, student attributes may emerge that can be used

as variables for tracking retention initiatives both at granular and institutional levels.

Manyanga et al. (2019) stated that few institutions truly have a retention agenda nor

agreed upon student attributes to apply as variables in these retention efforts. In various

retention models that have been discussed over the past few decades, Manyanga et al.

found that assessment and review at the course and program levels every 3 years should

be best practice. They continued and said that this best practice serves not only to inform

internal stakeholders of student satisfaction but to also reveal certain student attributes of

those who retain and can be applied as variables for tracking and predicting retention.

Premalatha (2019) agreed and stated, in her review of traditional outcome-based

education that included program review outcomes, that this model does not include a

process for identifying student attributes for retention. Premalatha continued that new

processes should be considered for outcome based education and program reviews to

include an intentional look at student attributes within these internal audits for student

success measurement and tracking for retention.

Higher education institutions have been faced with numerous problems ranging

from reduced student population, poor student performance, reduced funding and lack of

transparency in its operations (Baer & Norris, 2016). Krieb (2018) addressed this issue

from the poor student performance perspective. Krieb found that more study is needed to

identify additional student data for tracking retention to inform practice and policy overall

in higher education. Krieb brought a new awareness of the importance in identifying

student data for tracking retention at his institution, which is an outcome I hoped for this

research as well.

Tracking and Predicting Retention in Higher Education

Users of big data for retention tracking have shown inconsistency in its

application in higher education (Daniel, 2015). Daniel (2015) stated that this

inconsistency affects decisions made with respect to the experiences of the users,

institutional policies, and processes adopted by the institution. Daniel’s focus on big data

management in higher education emphasized identifying student data for tracking

initiatives as part of the big data management process for retention. Sperry (2015) stated

that there are valuable attributes to be considered in precollege student data when drafting

processes for retention initiatives. The importance of efficacious application practices for

student data in tracking and predicting student retention is evidenced by Niebel et al.

(2019) who said that some of the benefits of such practices are increased retention,

financial returns, and satisfaction of customers.

Tracking retention in higher education should consider the student body in its

entirety but not necessarily the identifying the same attributes applied as tracking and

predictive variables (Saunders et al., 2016). Saunders et al. (2016) indicated that subsets

of psychometric tools should be considered for retention tracking and potentially

predicting the retained student. They stated that this is not only complimentary to the

academic and demographic data that is commonly used in retention initiatives but also to

ascertain a students’ self-efficacy, confidence, and engagement for both in the classroom

and campus activities for connectedness. Xerri et al. (2018) agreed and stated that the

influence of student connectedness showed a direct relationship to the motivation to study

and therefore an increase in academic success that ultimately results in retention and

completion. Xerri et al. suggested that factors influencing this connectedness are largely

unknown and subsets of student psychosocial attributes would assist in not only gaining

this knowledge but further understanding the student journey in higher education. This

process for gathering subsets of psychometric or psychosocial data for tracking and

predicting retention was also suggested by Ganotice and King (2014). They stated that the

connectedness of students can potentially be measured from the perceived support of

faculty, parents, and peers—the most salient of these being that of the support of their

peers.

A current trend in higher education for student retention initiatives are high impact

practices (HIP; Provencher & Kassel, 2019). HIP are intended to bolster the support and

success of second year college students (Provencher & Kessel, 2019). They continued

and stated that early data has shown improved retention with students who are involved

with HIP. However, Provencher and Kessel (2019) were concerned that bias has occurred

with student selection for HIP and research should focus on institutions that shown

intentional equity of the HIP process, so retention data are not skewed. Murray (2015)

stated that HIP have shown increased student retention when considering student library

use and involvement. Murray suggested that student attributes for library use and

frequency be mined and used as retention tracking and predicting measures. In a similar

study, but a focus on asynchronous discussions boards for HIP, Perrotta (2020) stated that

this HIP also showed influence on increased student retention in the history program

being studied. Perrotta said that findings from this study set the stage for an increased

pursuit of student attributes from HIP and suggested that further research from qualitative

study would benefit.

Barriers to Data Mining

Lomet (2017) noted the importance of the protection of student data, the buy in of

faculty, and the cost effectiveness of any data mining tool for higher education. Bughlin

(2016) agreed and added that gathering big data for identifying variables requires a

system to mine and target an institution specific initiative but the barriers will be far

greater before the benefits can be realized. Nadasen and List (2017) stated that the

problem is the lack of information available to track or predict the retention of students

with a process for data mining and policy to inform best practices. Soares et al. (2016)

noted the lack of literature and research available to guide practice because much of the

literature is on Learning Management Systems (LMS), a specific academic group, or

small cohorts for identifying retention variables that only provides a narrow scope of

understanding in data mining. Wilderotter (2020) agreed and stated that a predictive

analytics model is needed for higher education that can be customized to specific

institutional needs while still underpinning best practice.

Important components to best practices in retention initiatives in higher education

are the attitudes and perceptions of administrators who have the responsibility for apt use,

defining, and application of, student data (Alsharari & Alshurideh, 2020). Staff and

faculty concerns of protected student information is another barrier to college

administrators’ application of data mining protocol for predictive analytics (Ekowo &

Palmer, 2016). Albalowi and Alhamed (2017) also addressed this concern, providing a

number of problems that have affected the adoption of data mining for student data

identification in higher education such as the culture or environment, absence of

appropriate infrastructure, and ethical issues related to the students.

Culture or Environment

Chaurasia et al. (2018) believed that user perceptions of analytics and data mining

tends to contribute to its acceptance and use as a medium for generating variables for

institutional research. They further added that many institutions do not have a culture of

data-based decision making because of these perceptions and therefore identification of

student data to assess learning as well as overall retention are problematic (Chaurasia et

al., 2018). Gagliardi et al. (2108) stated that the most significant challenges to the

adoption of data mining, specifically for variable identification used for internal research

in higher education, has been limited to perception of data managers toward the

institutions’ culture of this work. They continued by saying that data manager inability to

comprehend the benefits associated with the use of campus-wide analytics can adversely

affect an institution’s capacity to compete, perform their operation, develop a learning

culture, and retain students by potentially not wanting to share data, based on this

perceived data culture (Gagliardi et al., 2018). Cope and Kalantzis (2016) agreed and

stated that those higher education administrators who value introducing new methods of

thinking as well as a means of data-informed practice are typically met with barriers.

They went on to say that these barriers are the increasing cost associated with the

gathering, storage, analysis, and application of this data that is siloed and decentralized.

In addition to culture, Hadwater et al. (2018) stated that the barriers that have contributed

to the slow adoption of data mining efforts that identify student data for institutional

research include existing infrastructure, and institutional policies.

There is growing pressure for higher education to be a culture of evidence (Hora

et al., 2017). Hora et al. (2017) stated that data mining techniques applied to higher

education gives an institution the capabilities to improve institutional level operations

such as targeted recruitment for efficient admission operations for undergrad,

international, and graduate programming. Andrews and Lemons (2015) cautioned that

institutional level culture is not enough. They stated that for a decision-making culture

based on data evidence to occur, this practice needs to permeate to staff and faculty who

are responsible for the day-to-day services with students. Faculty still make decisions

based on personal experiences rather than data (Andres & Lemons, 2017). In any higher

education institution, there are multi levels of practice that feed into the culture of

learning (Klimek & Klimek, 2020). A seemingly monumental barrier to a productive shift

to a data driven culture will be the shared reality, vision, and trust of the IT department,

administrations, and faculty, said Klimek and Klimek (2020). They stated that for

dynamic development and dissemination of data technologies to occur, small steps with

data driven problem solving that require internal stakeholder integration can begin the

road to barrier removal.

Infrastructure

Cope and Kalantzis (2016) said that systems used by institutions of higher

learning do not support interoperable data analysis, suggesting that a wide range of data

for high level use and classroom data are stored in multiple online repositories and this

creates problems for narrowing down variables to identify as targets for internal research

efforts. Matsebula and Mnkandla (2017) noted that the absence of an appropriate

infrastructure specifically tailored for mining student attributes and behavior

characteristics in identifying variables has contributed to the inability for analytics to

support decision making in many facets of student success in higher education. Avella et

al. (2016) stated that student attributes can be collected, managed, and used to identify

variables for retention in many of the LMS purchased by institutions. However, Avella et

al. continued by saying that data mining from these LMS can be difficult. Albalowi and

Alhamed (2017) agreed but pointed out that although faculty pushback is the biggest

barrier, the benefits of the data mining capabilities in LMS can include improved student

placement, better enrollment rates, and enhanced attendance and academic warning

systems. Some institutions of higher learning have utilized certain LMS to evaluate the

quality of education offered, determine the enrollment rate, share profiles, acknowledge

the supported required in improving the learning experience, among other benefits

(Avella et al., 2016). Avella et al. continued by stating that these efforts are frequently

siloed and many still use manual data entry reporting such as excel sheet storage and

updates. Manual tracking systems and siloed data practices have proven to be ineffective

in resolving student retention in higher learning institutions (Matsebula & Mnkandla,

2017). Williamson (2017) acknowledged this and stated that educational data science has

predictive algorithms for retention in its grasp, but the level of expertise, manual tracking,

and available workforce can be a barrier. Williamson added that retention data should be

housed as a centralized data warehouse, but the responsibility of its use and application

should be widely dispersed with proper training and respect for protected student

information, as ethical issues are an ongoing concern.

Institutions of higher learning face new challenges relating to student information

uptake and internal analysis (Arendsdorf & Naylor-Tincknell, 2016), including

information related to global economics, political change, and ensuring the programs they

offer are relevant to market needs locally and nationally. Chaurasia and Rosin (2017)

added that, due to competition, institutions find themselves under immense pressure to

analyze and decipher large data sets, as doing so puts them ahead of the competition.

However, Stefanova and Kabakchieva (2017) contended that literature is inadequate to

support the knowledge base needed to apply data for retention efforts. Shein (2020)

agreed, stating that the knowledge base needed for data managers in higher education is

simply not in the literature.

Utilization of student data applied as performance indicators in tracking retention

in higher learning has been considered effective (Varouchas et al., 2018). Varouchas et al.

(2018) continued by stating that in higher education, data mining for student variable

identification for internal initiatives can be obtained from several sources such as social

media, information and learning systems used in the institution, and swipe card data. Data

mining processes such as these include a high velocity, volume, and variety of raw data

that is needed to be applied in tracking or predicting retention (Attaran et al., 2018). The

problem with this data mining is the lack of consistency, knowledge of data application

best practices, and sensical policy to align such endeavors (Varouchas et al., 2018). A

data mining model in educational settings is best used concurrently with data analytics

and the knowledge base required to understand the diversity, size, and speed of data

(Huselid & Minbaeva, 2019). The updates to the higher education data mining systems,

however, are long overdue (Tsai et al., 2015) and the research is not available for higher

education administrators to use as guides for retention tracking process and practice. Tsai

et al. (2015) continued by saying that management, processing, and application of

variables from large sets of data to track and predict success cannot be accomplished by

simple excel sheet formulas that were once considered traditional means. The collection,

management, and identification of variables for retention requires a more sophisticated

approach.

Ethical Issues

Roberts et al. (2016) indicated that the use of analytics tools in the higher

education sector has outpaced the ethical aspects associated with its usage. From the

perspective of higher education, data analysis related to students has comprised of

demographic details, enrollment survey results, student course assessment, the use of

library facilities, the academic performance of students, among others (de Freitas et al.,

2015). Elhassan and Klett (2018) suggested that an ethical perspective is needed to

consider student’s participation in decision making in order to promote a healthy

institutional climate so unbiased student data is being gathered from student attributes and

behavior characteristics needed for identification of variables for retention by the

administrators.

Ethical concerns related to data mining for student data is captured in the Roberts

et al. (2016) study. They indicated that the use of analytics tools in the higher education

sector has outpaced the ethical aspects associated with its usage. Roberts et al. continued

that the nonexistence of the students’ voices regarding the use of data mining has

presented challenges related to the acceptability of student attributes and behavior

characteristics use for identifying variables for further study. From the perspective of

higher education, data analysis related to students has comprised of demographic details,

enrollment survey results, student course assessment, the use of library facilities, the

academic performance of students, among others (de Freitas et al., 2015). Elhassan and

Klett (2018) suggested that an ethical perspective is needed to consider student

participation in decision making in order to promote a healthy institutional climate so

unbiased student data are being gathered from student attributes and behavior

characteristics needed for identification of variables for retention by the administrators.

Further, Lacerenza et al. (2018) stated that successful teams produce desired outcomes

but it is critical that team members demonstrate effective processes to achieve these

outcomes. They continued and stated that team development interventions are salient to

team survival but adherence to data policy and ethics is salient to institutional survival.

Summary and Conclusions

Major themes in the literature were discussed in relation to the key concepts of

defining retention in higher education, applying and identifying student attributes for

institutional research, tracking and predicting retention in higher education, barriers to

data mining. Defining retention in higher education can be both generalized and

institution specific (Dewberry & Jackson, 2018) depending upon the college mission and

accrediting body expectations (Conrad & Wilson, 1985). Tight (2020) stated that higher

education is missing an opportunity in tracking retention by not considering attributes to

apply as variables directly related to student external commitments.

What is Known

What is known about tracking and predicting retention in higher education is that

there is an overabundance of large unstructured data that needs to be disaggregated to

find the nuggets of significance for internal research (Huselid & Minbaeva, 2019; Shein,

2020) and the old system of manually gathering data from siloed data systems that do not

integrate, further increases institutional instability (Soares et al., 2016). Research

literature yields plenty of data mining options for higher education in regard to LMS

(Krieb, 2018) but these studies focus on variables that only serve academic purposes and

not those that inform policy and practice on campus wide data mining integration,

experiences and knowledge base of data administrators, nor the process for identifying

student data in tracking or predicting retention (Avella et al., 2016; Baer & Norris, 2016).

What is known about identifying student data from attributes and behavior

characteristics that are needed for tracking and predicting retention can be obtained from

a number of sources such as social media, information, and learning systems used in the

institution, and swipe card data among others (Varouchas et al., 2018). However, the

problem with this data mining is the lack of consistency, knowledge of data application

best practices, and sensical policy to align these undertakings (Chaurasia et al., 2018).

This all plays into the ethical dilemma that poses barriers in higher education if aligned

policies for best practices are not transparent to the staff, faculty, and students regarding

the use, storage, and outcomes of protected data (Elhassan & Klett, 2018).

What is not Known

What is not known in the areas of student variable identification and its place in

the efficacious tracking and predicting of retention in higher education are the perceptions

and experiences of the data administrators to this process that would directly inform

policy and practice (Gagliardi et al., 2018; Hadwater et al., 2019). Further, data managers

inability to comprehend the benefits associated with the use of campus wide analytics can

adversely affect the institutions’ capacity to develop a learning culture and retain students

by potentially not wanting to share data, based on a perceived data culture that may exist

or not (Hadwater et al., 2019). Higher education leaders need to know that the actual

barriers that may exist given the data manager knowledge base and a culture of data

support or not have contributed to the slow adoption of data mining efforts that identify

student data for institutional research (Ekowo & Palmer, 2016). These perceptions and

experiences from data administrators in higher education can prove to be valuable in the

evolution of data analytics for institutional stability (Albalowi &

Alhamed, 2017).

Gap in the Literature

This study seeks to fill the gap in literature that is needed to inform the practice of

identifying student data for tracking and predicting retention. Further, the outcomes of

this study aim to fill the gap in the literature regarding what is not known regarding the

process of data mining for student data, and that is the experiences and perceptions of the

college administrators responsible for these efforts. I also seek to have similar results as

Krieb (2018) in bringing awareness and understanding to those administrators in higher

education needing research results to inform a gap in the practice for identifying student

data in tracking and predicting retention. By accomplishing these objectives, this current

study will fill this informational research gap, thus benefiting future researchers who may

desire to explore and add on this topic.

Transition

The next chapter will detail the research design and method as well the process for

participant selection. Further, in Chapter 3, a discussion of the internal and external

validity will be explored, the role of the researcher, and data analysis plan and treatment

of the research data. A final summary will unify the information and complete the chapter.

Chapter 3: Research Method

The purpose of this basic qualitative study is to explore and describe perceptions

and experiences of administrators of 4-year colleges and their identification and

application of student data for tracking and predicting retention. In this chapter, the role

of the researcher as well as methods, design, and validity of research is discussed in

relation to the purpose of this study. A detailed description of participant recruitment and

selection strategies is discussed as well as the data analysis plan. The data analysis plan

includes a discussion of internal and external validity as well as adherence to ethical

procedures. Chapter 3 ends with a summary and segue into Chapter 4.

Research Design and Rationale

I used a bbasic qualitative research design to address the research question: What

are the perceptions and experiences of higher education administrators in terms of the

application of student data targeted for tracking and predicting retention? The central

phenomenon of the study is the application of student data for tracking and predicting

retention and how this process is perceived or experienced by administrators responsible

for such responsibilities at 4-year institutions. I wanted to make sense of the data analysis

process and inform practice in higher education regarding how student data may be

identified and influence retention initiatives. According to Barbour (2014), researchers

should confer with participants by asking questions and inferring meaning from

responses.

The basic qualitative research design is appropriate for this study because the RQ

is broad and open to unexpected findings. Knapp (2017) said the basic qualitative study is

appropriate to explore experiences of participants with knowledge of the topic and ascribe

meaning to those experiences. The basic qualitative design is also appropriate to explore

college administrators’ perceptions about the phenomenon of efficacious application of

student data for tracking and predicting retention. This design allows participants to

provide thick and rich accounts of their experiences and perceptions (Korstjen & Moser,

2017; Merriam 2015). Also, basic qualitative research is used to examine natural

circumstances in which individuals function to provide a practical understanding of real-

world problems (Korstjen & Moser, 2018).

Role of the Researcher

With the basic qualitative research design, the role of the researcher is to be the

instrument for gathering and analyzing data (Silverman, 2016). Qualitative researchers

influence the research process because study participants interact with researchers. In this

study, the interview process allowed relationships to be built. This influences the research

process and its findings, which is why it is important for me to be transparent about my

perspectives as the researcher and explicitly acknowledge any subjectivity.

My professional role is director of retention and completion at a 4-year private

nonprofit university. Inherent biases can threaten the trustworthiness of the study if not

kept in check. Shufutinsky (2020) suggested self-checking, which includes reflection,

feedback, and mindful consideration during qualitative research. Shufutinsky said

qualitative research is generally rooted in interpretivism, and therefore, the researcher is

responsible for the interpretation of participant responses. I used bracketing and epoche

methods to mitigate preconceptions and presuppositions that could taint the research.

Epoche practices are those practices that allow for active suspension of assumed

understandings by journaling throughout the research process and note taking during

data collection (Shufitinsky, 2010). Bracketing and epoche are often used interchangeably

in practice (Butler, 2016). The researcher, as the main instrument of the research, must be

constantly conscious of internal ideas, perceptions, values, prejudgments, and

connections to the topic under study (Creswell, 2013). Journaling throughout the study

process in its entirety and reflective note taking during data collection helps mitigate

research bias through awareness of self and processes.

As a qualitative researcher, I employed empathy as well as distance. Empathy

entailed putting myself into participants’ situations in order to better understand their

intent and meaning. Distance involves necessary awareness of my own values, which can

negatively influence data collection, and as the researcher, I must remain nonjudgmental

and nondirective. I did not have any known relationships with participants as coworker or

supervisor. This removes any potential power relationship with participants. No

information was purposefully omitted or altered. I acknowledged the importance of being

aware regarding my role during processes of collecting, analyzing, and interpreting data,

as well as mitigating preconceived biases by conducting a test run of interviews. These

efforts mitigate bias.

Methodology

This section includes procedures for coding and analyzing data, as well as

methods to ensure trustworthiness and ethics in research.

Participant Selection

The participant sample consisted of approximately 10 administrators from 4-year

U.S. colleges. Sample size result in qualitative research is determined by data saturation

(Etikan, 2016; Guetterman, 2015). Sample size for qualitative studies can vary when

using interviews for data collection. To recognize that data are saturated, some analysis

must occur during data collection. The researcher must notice when participant responses

become repetitive. I estimated that no more than 12 participants were needed to reach

data saturation; this was determined during the interview process. According to Patton

(2015), it is an acceptable practice in qualitative research to check with interviewees for

more information to enrich or clarify data to meet data saturation. If saturation is reached,

there is no need to seek additional interviews because themes, patterns, and concepts

repeat, with no new information being collected (Merriam & Tisdell, 2016). Selecting an

appropriate number of participants for a study leads to manageability of the data results,

richness of data, and relevant participant characteristics.

Chosen participants had decision-making responsibilities in their institutions.

Participants for the study were chosen via criterion-based selection, a form of

homogenous purposive sampling. Participants needed to have student data identification

or tracking and predicting of retention initiatives among their responsibilities. They were

identified through institutional public websites that list higher education administrator

profiles and contact information. The first 10 individuals identified as fitting roles and

criteria for participation in this study were selected and sent recruitment letters via email.

Email addresses were retrieved from public websites that were randomly chosen.

What follows are procedures for how participants were identified, contacted, and

recruited. An IRB approved letter of consent was sent via email. This consent outlined the

intent of the study, participants identity protection, treatment of data, and statement of

voluntary participation. Those who agreed completely filled out and signed the consent

form and returned it to my personal Gmail address listed in the consent form. Depending

upon participant location, the interview choices consisted of face to face, virtual, or

phone. Given the 30-day time frame that was presented in the introductory request letter,

the participants specified the best days and times for the interview and the preferred

format. The format of semistructured interviews are flexible and versatile, making them a

popular choice for collecting qualitative data (Kallio et al., 2016). They are a

conversation in which the researcher knows what she/he wants to cover and has a set of

questions and a foundation of knowledge to help guide the exchange and can be done face

to face, phone or virtual, depending upon research and participant preference (Magaldi &

Berler, 2018). Given the current COVID-19 restrictions the goal was to create a safe

space so the participant felt comfortable to reflect upon his or her own personal

experiences while maintaining social distancing protocol (CDC, 2020).

Instrumentation

Data collection was achieved through researcher-produced questions for

semistructured interviews. The questions found in the Appendix A, are based on the

literature review conducted relating to data mining practices for student variable

applications in tracking and predicting retention in higher education, as well as the

conceptual framework models chosen for this study. Also, in designing queries for

interviews, I was sure to use language that most participants are likely to understand

(Rubin & Rubin, 2012). Literature and the conceptual framework for this study guides the

development of the research question, but it was important to develop and use open ended

interviewing methods to avoid guiding participants’ answers (Rubin & Rubin, 2012). I

started with what Rubin and Rubin (2012) call a tour question that has participants talk

about their broad activities in their administrator role before asking about specific

experiences in student data applications in retention initiatives. The tour question is

intended to open a frame of context for participants to consider in a broad sense, while

the accompanying probes are worded to promote confirmation, clarification, sequence,

continuation, elaboration, and credibility (Magaldi & Berler, 2018). In this case, the

questions for phone, virtual, or face to face interviews are appropriate for qualitative

research given its fluidity and participants’ ability to elaborate on their answers

(Brinkmann, 2016). Each interview will last for approximately 60 minutes to provide

each participant with enough time to express and elaborate on each question (Magaldi &

Berler, 2018).

Basis for Instrument Development

The semistructured interview questions were created by me and follow protocol

from previous qualitative studies (Castillo-Montoya, 2016; Kallio et al., 2016; Magaldi &

Berler, 2018). Kallio et al. (2016) used a protocol for semistructured qualitative

interviews that included five phases: (a) identifying the prerequisites for using

semistructured interviews, (b) retrieving and using previous knowledge, (c) formulating

the preliminary semistructured interview guide, (d) testing the guide, and (e) presenting

the complete semistructured interview guide. I referred to these steps during the drafting

of the semistructured interview questions.

In a similar manner of study, Castillo-Montoya (2016) developed the interview

protocol refinement (IPR) framework for the development of an interview protocol. The

IPR method aims to support efforts in reinforcing the reliability of interview protocols in

qualitative research. The framework includes, constructing the interview questions and

ensuring interview questions align with research questions, has been completed (see

Appendix A). In this study, the interview protocol was used to address the RQ. To guide

my creation and alignment of the interview protocol, I used the Types of Interview

Questions from the IPR framework (Table 2). I also used the Types of Interview

Questions table as a reference to ensure internal validity (Castillo-Montoya, 2016).

Referencing this process ensured clarity, focus, and sufficiency of the questions to prompt

accurate responses that are aligned to the RQ.

Table 2

Types of Interview Questions

Types

Explanation of Type

Introductory

Questions that are relatively neutral eliciting general and

nonintrusive information and that are not threatening

Transition

Questions that that link the introductory questions to the key

questions to be asked

Key

Questions that are most related to the research questions and

purpose of the study

Closing

Questions that are easy to answer and provide opportunity for

closure

Note. From “Preparing for Interview Research: The Interview Protocol Refinement

Framework” by M. Castillo-Montoya, 2016, The Qualitative Report, 21, p. 823.

Magaldi and Berler (2018) said interviews are an accessible, affordable, and

effective method to understand a phenomenon in the world of research. Their approach

suggests that semistructured queries are an interpretive framework where the data

collected is not viewed as evidence of the truth or reality of an experience but rather a

context-bound and subjective insight from the participants. In this way, Magaldi and

Berler suggested that the researcher needs to be open to new insights and to honor the

participant’s experience in data collection by using the basic qualitative design to abut

this method that is exploratory in nature and permits the collection of rich data which can

answer questions about which little is already known.

Procedures for Recruitment, Participation, and Data Collection

Once IRB approval is received for this study, a criteria-based sample was

recruited to yield approximately than 12 administrators from a 4-year, U.S. institution.

Potential participants were identified through a higher education random google search of

4-year institutions’ public websites. Email correspondence to those academic leadership

having contact information listed on their institutions’ public site will commence to those

who have a job title, job responsibilities, or job description that involve data mining,

student variable identification and application in research initiatives, and participation in

tracking and predicting retention on any level at their institution. This initial email will

contain the details of the study on the 30-day timelines for the interviews, volunteer

consent, treatment of the data, confidentiality of the participants, and the purpose of this

study.

A minimum of 50 email invitations were sent for the first round of recruitment to

seek the approximately 12 participants. Return emails of interest from potential

participants will yield a self-identification for meeting the approved criteria. This will

continue until approximately 12 participants are identified. Each week an additional 25

institutions were googled for participants meeting criterial and contact information

displayed on their institution’s public site. This continued each week until the minimum

number of 10 but no more than 12 participants meeting the criteria and accepting the

terms is met. Participants who returned the emails with interest received the letter of

informed consent within 24 hours to review and sign. Any participant that showed interest

but did not respond were sent two additional follow up emails to confirm their interest or

not. Should there have been participants who were unwilling for any reason to continue in

the study at any time were told they may contact me via my listed email and I would have

removed them from the study.

All interviews were recorded with the Google transcribing tool that is a voice to

text software. The use of this software supports the need to ensure valid and reliable data

from the interviews as well as have a cross checking system for me to review for

accuracy (Creswell & Poth, 2016; Given, 2008). This process was performed while

keeping the participants’ identity protected and honor the confidentiality of the

participants, in case they would inadvertently provide identifying information and need to

be redacted. The interviewees were informed of potential follow up, within 2 weeks after

the interview, via email if further clarification is needed. At that time, all participants will

receive an email stating that the interviews are completed, and the analysis has begun.

Any participant interested in viewing the final study can email a response and a copy was

emailed upon completion.

Data Analysis Plan

What follows are the data analysis approaches for this study. I used Feng and

Behar-Horstein’s five-step procedure to analyze the data collected for the study:

1. Step one is data cleaning and participant coding alignment to responses in an

excel sheet format or manual first round transcript coding.

2. Excel sheet data or direct transcript was imported to NVivo software system.

3. Word frequency analysis will then be conducted through NVivo word

frequency query feature.

4. Text coding and reference extracting includes the text search query feature to

identify the most frequently occurring words to code the responses content of

all sentences or paragraphs for each participant within each question. NVivo

refers to these words as nodes. Nodes are also known as categories in the

qualitative induction coding process. Content that includes the most frequently

occurring words were identified as references of the nodes. The text coding

summary from NVivo shows the number of references for each node or

category.

5. Matrix coding, mind mapping, and data relationship queries for continued

inductive analysis allow comparisons across and between different nodes or

categories and references to categories within the participant responses for

themed focus.

Software programs can be useful in organizing large amounts of data and assist

the researcher with assigning codes to data (Korstjen & Moser, 2018). The use of

software can simplify the analysis process without sacrificing any significant meaning

found within the data. I used inductive coding to tag meanings in the perceptions and

experiences of student variable identification from participants (Merriam & Tisdell,

2015). Given the probability of diverse responses, inductive coding was used as a cross

check to NVivo in the matrix coding and inductive analysis step of Feng and Behar-

Horstein’s data analysis approach. This will continue with each component of the RQ:

perceptions and experiences of student variable application to research, perceptions and

experiences of student variable identification in tracking and predicting retention, and

perceptions and experiences of student variable applications in tracking and predicting

retention. The bigger categories are the overarching themes while the subcategories are

theme supporters (Korstjen & Moser, 2017). NVivo software analysis uses the word

nodes for the subcategories and theme supporters as references to nodes that feed into the

major themes to be addressed in the RQ (Feng & Behar-Horstein, 2018; Welsh, 2002).

This is where the participants’ responses become a story from the data. The themes can

tell the same story from different perspectives, or several different stories that connect

with each other (Given, 2008). This final phase in analysis involves connecting the stories

through connecting themes in data with word clouds, frequencies, percentages, or tables

to finalize the analysis from NVivo and researcher notes (Babchuk, 2017).

Once the data are collected, I analyzed it through the matrix coding approach. I

also employed an iterative process throughout the data analysis process for the purposes

of organizing and managing the data (Babchuk, 2017; Given, 2008; Merriam, 2015). This

process will involve labeling interview notes, transcripts, and participants with

confidential identifiers. This process included organizing key elements of the data

relevant to this study, including a priori and axial coding systems (Babchuk, 2017;

Merriam, 2015). The a priori codes were created from the conceptual framework

supporting this study from Attarran et al. (2018), and Gandomi and Hader (2015). A priori

codes were highlighted in the transcripts as key words and phrases. During the open

coding procedure of the study, I used repetitive words and phrases of meaning that

emerge from the data, as well as those seen as emphasized by the participants, (Korstjen

& Moser, 2017) to be designated as codes (Merriam & Tisdell, 2015). During the final,

axial coding I categorized codes into groups and identify patterns that will become major

themes (Lichtman, 2017). This final process formed the basis for my findings and

conclusions of this study (Creswell & Poth, 2016).

In basic qualitative studies a researcher must identify and assess discrepant data

(Levitt et al., 2017). Discrepant data is an occurrence that cannot be accounted for or

explained and can signal defects in the data (Korstjen & Moser, 2018). It was important

to examine and confirm discrepant data, the inconsistent pattern of data, with that of the

other resources and review with participants. Maxwell (2008) stated that distinction

between categories or themes may be the source of negative or discrepant data and can be

resolved with probing questions or follow up review with participants. Once this is

completed, the discrepant data will be shared in the analysis and findings of this study

and explained (Creswell & Poth. 2016).

Trustworthiness

Trustworthiness in qualitative research must demonstrated to show proper

methods and rigor were used throughout the data collection and analysis process

(Babchuk, 2017). Protocols in trustworthiness include demonstrating credibility,

transferability, dependability, and confirmability (Amankwaa, 2016; Korstjens & Moser,

2018). Maxwell (2008) stated that researcher bias is a thread throughout the research

process that provides a quality of awareness and that we should not suppress our primary

experiences. He continued and stated that, conversely, we do not allow this awareness and

related experiences to overwhelm nor drive the research process, but rather elevate

mindfulness and use it as part of the inquiry process. Lichtman (2017) stated that the

researcher is the conduit through which participant relationships are built to yield insight

in the data. Lichtman explained that self- reflection and subjectivity within the steps of

building participant comfortability in the interview process for data collection does not

cause a paradox or confusion. Rather, the researcher creates the awareness needed to

show the sense of self and therefore demonstrates deep understanding that directly

influences all aspects of trustworthiness in qualitative research (Lichtman, 2017).

Credibility

Credibility was authenticated in this study by assuring the participant response

was received as it was intended through member checking (Patton, 2014). As stated by

Korstjen and Moser (2018), I repeated probing questions, take side notes during the

interview, and sharing the tentative results of data analysis with the participants via email.

An agreement with Madill and Sullivan (2018) I used member checking to repeat the

interview questions with different tone or wording without changing the meaning. I

provided participants with the opportunity to restate their answers during the interview

and amend the meaning of their statements after the interview (Madill & Sullivan, 2018;

Patton, 2014). I did, during interviews, noted vague responses, repeated questions with

clarifications, and reflected participants statements back to them for clarification (Madill

& Sullivan, 2018).

Another credibility authentication method that I used is reflexivity. Korstjen and

Moser (2018) described this as the process of critical self-reflection about oneself as the

researcher and one’s biases, preferences, and presumptions as well as the relationship to

the participants and how this relationship may affect their answers to questions. As

Maxwell (2008) advised, I maintained a reflexive, internal credibility process during the

semistructured interviews. By journaling and taking notes throughout each interview, I

monitored my own explicit and implicit assumptions in all phases of this qualitative study

to enhance credibility (Maxwell, 2008).

A third process to confirm credibility in the study is how discrepant data was

handled. Negative or discrepant data is described by Patton (2014) as exceptions to the

patterns found in the data. When negative or discrepant data occurs, I reviewed and

reflected from a cross checking perspective via interview recordings, researcher memos,

and journal to determine useful support of the study as suggested by Bashir et al. (2008).

This allowed me to record the experiences of participants within and beyond the

immediate context (Korstjen & Moser, 2018). So, I continually checked and verified the

data processes to ensure that the results are robust, rather than a simple justification of

any assumed findings (Spiers et al., 2018).

Transferability

Transferability relates to the ability to transfer the results of the study to a

population differing from the one used in the data collection (Amankwaa, 2016; Babchuk,

2017; Kallio et al., 2016). Participants were selected from those serving in decision-

making positions in higher education from 4-year, U.S. institutions. Criteriabased

selection was used to recruit higher education administrators with specific knowledge in

data mining, student variable identification and application for internal research, and the

tracking and predicting of retention on any level at their institution. By providing this rich

description of the participants and the research process, the reader of this study can decide

whether the findings are transferrable to their setting (Korstjen & Moser, 2018). Korstjen

and Moser continued and stated that the reader, not the researcher, can make the

transferability judgment. The job of the researcher is providing as much context as

possible so lens can be clear for the reader to see applicability to their setting or not

(Arensdorf & Naylor-Tincknell, 2016).

Dependability

Dependability is also necessary in a basic qualitative study to show reliability of

the data collection and the analysis (Amankwaa, 2016; Babchuk, 2017; Kallio et al.,

2016). For this study, audit trails will be easily accessible in a few forms from the start of

this research study to the development and reporting of the findings as needed to show

transparency as stated by Korstjen and Moser (2018). Once such audit trail can be found

in the records of the research path that are to be kept throughout the study and 5 years

after its completion. I have this information password protected and kept at my home

office with desk drawer key that I alone can access. Another audit trail used in this study

was the a priori coding system protocol used by the researcher for the interview questions

and audio transcription for cross checking interview data of the participants as suggested

by Babchuk (2017). Finally, the handwritten reflexivity notes and journal will be

accessible as well as the recordings of the participant interviews for review as needed

(Silverman, 2016). This protocol follows a specific stage by stage process from general

memo taking and journaling through reflexivity, taking the priori coding in the interview

questions that can be further used during the data collection process through deductive

coding and cross checking with the NVivo software (Feng & Behar-Horenstein, 2019).

Confirmability

Confirmability refers to objectivity or the ability of others to confirm findings

(Stahl & King, 2020). Confirmability refers to the researcher’s transparency and

documentation of processes (Korstjen & Moser, 2018; Nowell et al., 2017). As Koch

(1994) recommended, I included markers, such as the reasons for analysis choices, so that

others can understand how and why analysis decisions were made. I enhanced

confirmability in this study by describing the process used for data collection and analysis

as suggested by Meadows (2003). I provide a detailed description of the sequence for

data collection, coding, and analysis to deliver a clear and well-defined accountability for

the process (Bochner, 2018; Meadows, 2003).

Ethical Procedures

I submitted a request for internal review board (IRB) approval before conducting

any research with human participants. This process is in place to hold accountability for

researchers. The IRB document contains questions that must be answered by the

researcher in regards to participant selection criteria, informed consent document, contact

intended for any vulnerable populations, instrumentation tool and how it was used, and

treatment of data. Once all research protocol met the standards for the protection of

participants, then approval was given (IRB Approval # 04-22-21-0672595). Upon IRB

approval the potential participants were contacted via email with informed consent that

includes the purpose of the study, type of data collection, and any risks if indicated.

Further, the informed consent clearly stated the voluntary nature of participation and

ability to withdraw at any time as well as complete confidentiality during and concluding

the process of the study. The study did not require the use of vulnerable populations.

Further, the participants of this study could leave at any time during the process. The

semistructured interview questions approved by the IRB are found in Appendix A. I did

not be conduct this research in my direct work environment nor with any higher

education professional known to me, so these rule out any potential conflict of interest or

power differentials. I have chosen not use participant incentives for this study.

Treatment of Data

All email correspondence taken place beginning with participant recruitment up

until data analysis will remain in a password protected hard drive. All transcribed audio

recorded data are kept within google docs account that only I can access and is password

protected. Any handwritten notes to support data in hard copy file format are housed in

my home office within a locked drawer and I am the sole key holder. Participant names

were coded to protect anonymity during data analysis and when delivering the study

findings. Confidentiality was maintained throughout the process by changing the names

of the participants and general references to 4-year schools without identifying name of

schools. It is required to keep all data and results for 5 years after the completion of the

study (Babchuk, 2017; Knapp, 2017). After the 5-year period is concluded all files will be

either shredded or deleted as indicated for hard copy and electronic copies of the research

study.

Summary

This chapter explored all that is involved in participant selection, recruitment, and

the data collection and analysis that comes from the instrumentation. Trustworthiness and

ethical procedures being followed for this study were also briefly discussed. Once

completed, this study will offer an account of the perceptions and experiences of

administrators of 4-year, U.S. colleges regarding the application and identification of

student data for tracking and predicting retention. Participants are selected based on certain

criteria needed in either their job title, job responsibilities, or job description. Recruitment

for at least 10 but no more than 12 participants began with a random Google search of all

4-year, U.S. colleges that have public websites. Within these websites, another search was

conducted to view any administrators that may meet the criteria for data mining, student

variable identification and application to internal research efforts, and predicting and

tracking retention on any level at their institution. Contact information posted allows for

email and phone contact to recruit participants. Protocol was followed for the IRB approved

informed consent and the researcher produced semistructured interview questions. The 30-

day timeline started the day the participant sent back the approved and signed informed

consent form. This 30-day timeline allowed for flexible dates and times for participants as

well as any time needed for brief follow up. In the next chapter, I reflected upon the details

of the study results through the discussion of the data

collection, analysis, and trustworthiness of this study.

Chapter 4: Results

The purpose of this basic qualitative study was to explore perceptions and

experiences of administrators of 4-year colleges in terms of their application of student

data for tracking and predicting retention. This study was driven by the RQ: What are the

perceptions and experiences of higher education administrators of 4-year colleges in

terms of the application of student data targeted for tracking and predicting retention?

Efforts to answer the research question are further described in this chapter. This chapter

includes the setting of this study as well as specific participant information. describe

minimum requirements for participation. A brief description of the data collection process

precedes the data analysis section that is the bulk of Chapter 4. A thematic coding

approach was used via NVivo. This coding process is discussed in detail in the data

analysis section. A brief summary of the chapter follows with a transition to Chapter 5.

Setting

Participants in this study were chosen from 4-year higher education institutions

that had either a direct responsibility for retention initiatives or were indirect supervisors

or committee members. Of the participants, four were from private and six were from

public nonprofit 4-year colleges. Four participants had direct experience with identifying

student data for retention initiatives, and it was in their job to do so as administrators.

Three participants were also directly involved with identifying student data for retention

initiatives via the nature of their direct student advising roles and specific student success

data, analysis, and outcomes. Three participants were indirectly involved in identifying

student data for retention initiatives via membership and supervision of a student service

team that was tasked to do these responsibilities in either academics, student support,

athletics.

This study was conducted during the COVID-19 pandemic. My focus was on

perceptions and experiences of participants regarding data for student retention initiatives.

Participants talked little about the pandemic and the influence of mandates on their

retention initiatives. This could have been an unexpected effect to consider during data

interpretation; however, during the inductive coding process, these items became

irrelevant to the results, as participants did not relate any of the interview questions to be

influenced by the pandemic.

Data Collection

The 10 semistructured interviews with participants currently employed in

administrative higher education roles served as sources of data. Each participant had

direct or indirect responsibilities for defining and selecting student data for retention

initiatives at their 4-year U.S. institution (see Table 3).

Table 3

States Represented by Study Participants

Number of

Participants

State

Indiana

Michigan

Missouri

Pennsylvania

Ohio

Each interview was manually coded for initial patterns in the interview data as

well as researcher notes and memos during interviews. Interviews were conducted

remotely and audio recorded and transcribed via the Google transcription tool. After

receiving consent from each participant, they were assigned a participant number to

safeguard confidentiality and privacy. Each participant engaged in one Zoom or Google

Duo meeting that lasted 45 to 60 minutes. The format was dependent upon participant

preferences given their preferred software platforms for virtual meetings. I encouraged

each participant to schedule meetings at their convenience where they could be either in

their own home or a private office if they chose their work environment. In doing so, each

participant would have only those around them they felt comfortable with. Before the

start of each interview, I reminded them of the consent form and meeting recording and

transcriptions, and thanked them for their time and commitment. Once interviews were

completed over the course of a 3-month period, I read each transcript thoroughly while

listening to audio recordings to correct any grammatical or inaccurate transcriptions

errors. There were no notable variations during this procedure. I then began the manual a

priori coding process. I then uploaded data to NVivo 12 for data reorganization to begin

data exploration. This process was the beginning of coding relevant information that

would generate themes to answer the RQ.

Initially, the number of participants sought was at least 12. However, data

saturation was reached after 10 participant interviews. Saturation occurred when no

additional data were found that was different from the first nine participants. As I saw

similar experiences and perceptions related to the RQ repeatedly, I became confident that

data related to answering the RQ was saturated. During data collection, I began to see

repeated patterns in transcripts during the sixth interview. However, in order to address

diversity of data and to make certain that saturation was based on the widest range of

data, , I continued with four more participants to confirm saturation.

Throughout the data collection process, the semistructured nature of the interview

tool allowed participants to lead the discussion and provide open ended answers for their

experiences and perceptions regarding the evolution of data mining at their institution.

Further, the open ended semistructured interview tool also allowed the participants to

describe their specific duties related to the retention initiatives and student variable

selection and defining processes.

Data Analysis

A framework analysis was conducted and consists of several stages such as

familiarization, identifying a thematic framework, coding, charting, mapping and

interpretation (Stahl & King, 2020). A framework analysis is used in qualitative research

when a naturalistic approach to data gathering is sought and the researcher seeks to

understand phenomena in context-specific settings. The real world setting allows the

participants to be comfortable in the research process and the researcher does not attempt

to manipulate the phenomenon of interest and only try to unveil the ultimate truth

(Bochner, 2018).

Data analysis began with the initial a priori coding (see Appendix C) of audio

transcripts. Each transcript was reviewed line-by-line for six a priori codes. Once initial a

priori coding was complete, I began open coding. I completed line-by-line manual open

coding of transcripts to determine additional codes found repeatedly throughout each

transcript. Audio transcripts with completed a priori and open manual coding of all 10

interviews were then uploaded to NVivo 12. Each was loaded and labeled as a case file

with designated participant confidential identifiers. Each case file was manually coded

within NVivo to create a second coding process of interviews. This helped the analysis

process remain consistent in terms of emphasizing key points during coding that were

cross-checked with researcher notes and memos during the open coding process. This

resulted in 811 initial codes initially. Some of the most common codes, found five or

more times in transcripts were: not data informed, no communication among programs

and departments, lack of defined processes for retention initiatives, student attributes

changing from term to term, lack of leadership driving best practices for student

retention, lack of knowledge of who is responsible for identifying student attributes and

variables for research, faculty and staff frustration, lack of centralized data warehouse

platform, lack of data mining knowledge, lack of people to gather data for student

retention, reactive rather than proactive decisions, lack of internal-external resources for

analytics software, and lack of transparency for student retention initiatives.

During the selective coding phase, I searched to find categories emerging from

similarities in open codes. Using word-mapping and relationship types queries in the

software, I took all the vignettes and the open codes and mapped them into a tree-map

and word cloud. Diverging instances of the identified patterns and trends were noted from

the narratives of the participants and they gave new meanings to my understandings of

the text. Some coding patterns found most common, five or more times in the transcripts,

in this selective coding process were: inconsistent data tracking, no common data mining

definitions, no common data mining process, decentralized student retention initiatives,

decentralized data retrieval, manual raw data kept in excel and google docs, too much

time to gather and analyze data, and decisions made before data can be presented. Figure

2 includes the summary of the data and analysis process for a priori, open, selective, and

categorical coding. Using NVivo 12 software, I continued with word-count queries and

relationship or connection types cross checking tool as another means in discovering

selective codes from the data.

Figure 2

Data Analysis Process

In analyzing the depth of codes, or the quantity of vignettes assigned to a group of

codes, or grouping of open codes, selective codes emerged from the data. For the

purposes of this study, the researcher defined depth as having 10 or more vignettes

assigned to a code. Thematic coding resulted from the connections both within and across

the open codes and selective codes. Connections across the selective codes were analyzed

with the mind-map tool within NVivo. When building the mind-map, each time a vignette

linked directly to a code, I reviewed that vignette for connections with other codes and

each had a designated color assigned to it. If there was a connection, NVivo connected the

codes with that designated color coded line. The selective codes with the most

connections formed the start of thematic coding.

Open

•

Each line of transcribed interview text was coded line by line manually

•

Each vignette from manual coding was entered into NVivo and either coded with a unique new

open code or linked to an existing open code

Selective

•

Mind

mapping software was used to group open codes into categories. All vignettes were

transferred into the wordmap, linked to an open or selective code

•

NVivo word

counts of transcribed interviews were used as second check for additional codes or

categories

Categorical

•

Word

mapping software was used to help discover themes by linking codes and vignettes from

open and selective coding where a direct relationship was clear from the NVivo "relationship

types" queries

•

Selective codes with the most relationships formed the foundation for coding into categories

•

This same process continued in the final coding phase for themes, but only the categories were

queried for word count and relationship types

Emerging Selective Codes into Formed Categories

Inductive and comparative strategies endorsed by Merriam and Tisdell were used

to analyze similar data revealed in multiple codes that grouped into categories, and then

into final themes. The initial coding cycle of the interview transcripts resulted in 23 initial

relationships from mind map coding. In subsequent reiterations my list of categories was

narrowed to nine. I consolidated several items into similar threads following Saldaña's

recommendations of sorting and shifting coded materials into categories, the relationship

between variables, patterns, and themes. The clarity and depth shared in the interviews

generated a total of 464 coded subdivisions during my analysis (Table 4). Discrepant

cases that were found in the data were those that were misaligned from the majority of the

interviewee results. Discrepant cases were set aside and used in a brief discussion in each

themed area to be deliberated in the Results section of this chapter.

Table 4

Frequency of Codes to Categories and Subthemes

Categories/Subthemes Total Word from Coded Vignettes/ Frequency

Interview Participant Researcher Count Total Segments Final Round

Relationships Transcript Segment Count Notes Segment Code

Siloed Data 79 A1 24 15 39

Siloed Communications 71 B2 29 11 40

Time Consuming 66 C3 36 18 54

Manual data retrieval 59 D4 33 13 46

No centralized data warehouse 53 E5 39 11 50

No common data language 42 F6 35 19 54

Inconsistent data practices 34 G7 26 17 43

Varied student attributes 31 H8 32 11 43

Partial data driven decisions 29 I9 27 21 48

J10 35 12 47

Total: 464 Totals: 316 148 464

Results

In this section, I discussed the main themes associated with answering the

research question and any subthemes related. The research question guiding this study

was: What are the perceptions and experiences of higher education administrators of

4year colleges in the application of student data targeted for tracking and predicting

retention? Themes that emerged were:

Theme 1: There are no common data mining practices or definitions;

Theme 2: Student retention decisions are only partially driven by data; and

Theme 3: Data is siloed and subthemes are (a) manual data retrieval processes and

(b) time consuming data analysis process.

Theme 1: No Common Data Mining Processes or Definitions

According to the perspectives shared by the college administrators interviewed for

this study, there is frustration in understanding the data mining processes for student

retention initiatives. Further, the college administrators were frustrated by the lack of a

common language or definitions at their institution for the data analysis process and the

variables used for research. The participants also shared that although there were common

academic and administrative language that is understood by most faculty and staff, this

did not translate over to the data collection, storage, and analysis that leads to results of

unknown origin.

Regarding Theme 1, D4 stated:

Honestly, I have no idea what student data are used for retention initiatives….the

problem is that we have no common data language. The other problem is that no

one seems to know how we get the data results or where they come from.

I9 said:

This is a frustrating time for us because of COVID-19. We have to know what our

students are thinking and how they are feeling. But defining the parameters for

measuring this is just as difficult as understanding how to do it.

Participant F6 was frustrated with leaderships and in regard to Theme 1 stated:

“Transparency shouldn’t be a privilege in student retention. We should have a common

language and understanding for the process and where the data originates.”

Two discrepant cases in regard to Theme 1 came up in the data analysis process.

C3 felt that defining a common language for gathering and presenting data for student

retention was important, but not as important as trusting the leadership put in place to

track and monitor student retention. C3 further stated:

I have an important job and it matters to a great many people that we are driving

best practice and student service delivery with data. However, they do not ask

how we define the data or the outcomes. Faculty and staff just want to know what

to do to retain students. It comes from trust.

J10 stated:

Data for retention initiatives is going to vary for each institution as well as the

definitions for the variables and research process. It is more important to focus on

pulling the right variables and attributes to get the outcomes we need to drive our

practice. That in and of itself is hard enough.

Theme 2: Student Retention Decisions are Only Partially Driven by Data

Most of the administrators interviewed expressed concerns of not being a data

driven institution. This concern played out in multiple experiences that were shared

during the interview process. These concerns were related to being reactive rather than

proactive with student retention and leadership lacking explanation for policy changes.

The participants also shared great frustration with leadership adding or removing services

based upon what other institutions are doing or trends, rather than having a reliable data

analytics structure for predictions related to student retention.

A1 spoke of these lack of data informed decisions at their institution and stated:

We do not have a choice in high level student retention reporting. We have to

have strong, reliable data for reporting to IPEDS and HLC accrediting. The

department and program level data are hard to find, collect, and analyze to make

decisions on student preferences for format delivery and types of services needed.

We rely on some student survey outcomes that are around 15% participation, and

the rest is gut instinct and results of many discussions and meetings.

B2 also shared concerns at their college on the lack of data informed decision

making and the toll it has taken on retention rates. B2 stated:

I know COVID-19 plays a factor in declining retention rates, but ours were falling

long before this. We cannot make a data informed decision in one area like

admissions and lack the wherewithal in other areas just as important, like student

retention.

H8 shared more detail about the frustrations of higher education decision making

downfalls. H8 stated:

We are supposed to be here to help our students succeed. We cannot fail at this.

But how are we supposed to help our students if we fail to help ourselves? We

should see through the data outcomes lens, what are students need and what the

level of need is so we can act appropriately. But, our institution continues to be

impatient with data analysis and invests in poor time management, rash decisions,

and lack of resources to influence positive change in student retention.

One discrepant case was noted. G7 stated:

We are moving toward a data informed decision-making culture. It has taken

years to do so, but we are less and less desiloed [sic] and more and more

integrated with communication, data analysis, and working together for student

retention efforts. Although we lack a centralized data warehouse, we make up for

it in intentionality of our manual data retrieval processes.

Theme 3: Data is Siloed

All participant data resulted in solidarity on this third and final theme. Although

there were discrepant cases in the subthemes, the misalignment was slight, but enough to

discuss as discrepant. The college administrators viewed higher education in general as

competitive based on fear. This fear is from lack of resources and a fickle climate for

degree seeking consumers. The data analysis from the college administrators made clear

that internal fear at individual institutions existed for the same reason and created the

siloed effect. Each department and program have a solitary mission to grab all they can

and seek the attention of Trustees and Executive leadership to survive, while subtly

driving down the success of other departments and programs. Sharing data and

integrating efforts to increase student success is not a priority, unless it serves to advance

the success and presence of the department itself.

E5 said:

The competition amongst ourselves makes my job difficult, if not impossible. We

are siloed as it is and gathering data for student retention initiatives is like herding

cats. Some programs will gladly give help as well as ask for help with deciphering

data and choosing student data to research. Other programs will not budge and

they assume that any data they gather belongs to them, but it belongs to our

institution. This siloed mentality will eventually close our doors.

D4 said: “Retention initiatives would be much easier if I didn’t have to gather

from 11 different islands. Each island has their own government and rules. None of them

work together.”

H8 said: “Our customer is our student, bottom line. If that is not a good enough

reason to come out of your fortified bunkers to share information and keep our doors

open, then what will it take?”

A1 agreed and said:

We have to be able to gather data and make decisions in unison. The internal

communication breakdown and data corruption will only serve to hurt ourselves

and our students. We have to coexist, and the common ground is institutional

survival through student success and retention.

Subtheme A: Manual Data Retrieval Processes

Many of the participants described data collection and storage as a manual process

held within Excel sheets and Google docs. The data is raw and not aggregated until

someone needs to access an aggregate response to a question for driving practice or

policy. There is then a continuation of this manual process through running PIV tables,

formulas, and a series of cut and paste activities to try and understand the data and the

variables that speak into it. There is a large margin of human error as well as inconsistent

process and unreliable data outcomes.

D4 said:

You cannot move the needle on retention if you do not know where the needle is.

We do not have one haystack to find the needle, we have many haystacks. It

equates to grabbing hay handful by handful and potentially having to do it again

and again because you missed a few and still cannot find the needle.

H8 said:

We are missing a lot of opportunities here and we need get caught up. Keeping

data on spreadsheets in dozens of offices across the institution is not only a

HIPPA violation in some cases, it is not best practice. Manually retrieving data

from department to department reaps only muddied results. We need a better way.

Participant J10 in this study shared similar sentiments on this issue and stated:

Right now, in my job, I have to track 421 students’ success. This is all on Excel

spreadsheets and I have trained myself on how to run certain formulas and PIV

tables to understand the needs of my students. There is not data analytics software

and we all need it desperately.

However, G7 said:

We have a mix of some department level data analytics software but still rely

upon some manual data processes. The software helps speed up the deciphering of

raw data and decreases human error as long as the human conducts data input

correctly. The need for manual processes may never go away. This may be a

necessary inconvenience for the sake of cross checking if nothing else.

Subtheme B: Time Consuming Data Analysis Process

College administrator experiences have been described as not enough time to

process data because it takes a lot of time to process data. My analysis of the transcripts

gleaned perceptions that spoke of unrealistic expectations for data results in days that, in

reality, take weeks to clean, organize, and decipher. Even the best of data analytics

software takes hours to input data points, choose filters, and run reports. This can be an

iterative process depending upon the software and the task at hand. But to input data into

columns of Excel sheets and Google docs that require organizing, filtering, formulas.

And PIV tables just to get started is a whole different stress on time and human capital.

E5said:

There are days that I want to give up. So much time is invested in gathering sheets

of data to organize, clean, and start to run PIV [sic]. Then someone changes the

request or deadline in the slightest and panic ensues. It shouldn’t take days and

days to get data analysis to work for us.

B2 stated:

The time it takes for me to produce even the simplest of data requests, such as a

course roster showing attendance concerns, takes a day or two. This is because it

has to be cross checked with the registrar because the LMS and SIS do not talk to

each other. I could have helped at least 6 to 10 students one on one in that time.

F6 said:

There is not enough time in any given day as it is. The process to grabbing raw

data from excel sheets and playing with filters and rows and columns is just

exhausting. Part of the time issue is just from using the help function on Excel to

learn how to do a filter or modify a PIV table. Our institution has to invest in

analytics software. We need it for many reasons. But just to be able to give back

hours upon hours of my time and that of my team would be invaluable.

I9 stated: “Time is not the issue, it is the lack of resources. Getting good data takes

time, that’s the nature of the beast. We have to do it right or it is not worth doing it at all.”

This section went through the three themes of the study, with the third theme

having two subthemes. Each area gave a description of any discrepant cases except for

Theme 3. The rest of this chapter will speak to the trustworthiness of my study and then

provide a brief summary.

Evidence of Trustworthiness

Trustworthiness

Trustworthiness in qualitative research must be demonstrated to show proper

methods and rigor were used throughout the data collection and analysis process

(Babchuk, 2017). Protocols in trustworthiness include demonstrating credibility,

transferability, dependability, and confirmability (Amankwaa, 2016; Korstjens & Moser,

2018). I followed the guidelines provided by Walden University’s Internal Review Board

and the research recommendations shared in the literature from Chapter 3 to ensure I

executed my study with rigor and transparency to the processes contained within it.

Credibility

Credibility is authenticated in this study by assuring the participant response was

received as it was intended through member checking (Patton, 2014). I provided

participants with the opportunity to restate their answers during the interview and amend

the meaning of their statements during the interview if needed (Madill & Sullivan, 2018;

Patton, 2014). During interviews I addressed vague responses with repeating questions in

order to reflect participants statements back to them for clarification (Madill & Sullivan,

2018). There were instances where participants were unsure of the meaning of the interview

question and I repeated the interview question by changing the tone or highlight certain

words without changing the meaning.

The second credibility authentication method I used was reflexivity. Korstjen and

Moser (2018) described this as the process of critical self-reflection about oneself as the

researcher and one’s biases, preferences, and presumptions as well as the relationship to

the participants and how this relationship may affect their answers to questions. By

journaling and taking notes throughout each interview, I monitored my own explicit and

implicit assumptions and values in all phases of this study. Finally, confirm credibility in

my study I used discrepant data findings in each theme described in this chapter.

Negative or discrepant data is described by Patton (2014) as exceptions to the patterns

found in the data. When discrepant data was found in participant transcripts, I reviewed

and reflected from a cross checking perspective via interview recordings, researcher

memos, and NVivo data queries. This is done to determine useful support of the study as

well as responses not aligned to the final thematic results in order to show rigor and

transparency.

Transferability

As stated in Chapter 3, transferability relates to the ability to transfer the results of

the study to a population differing from the one used in the data collection (Amankwaa,

2016; Babchuk, 2017; Kallio et al., 2016). By providing this rich description of the

participants, setting, sample size, the research process, and the findings (Korstjen &

Moser, 2018). The reader of this study can decide whether the findings are transferrable

to their setting. The job of the researcher is providing as much context as possible so lens

can be clear for the reader to see applicability to their setting or not (Arensdorf & Naylor-

Tincknell, 2016).

Dependability

Dependability is necessary in a research study to show reliability of the data

collection and the analysis (Amankwaa, 2016; Babchuk, 2017; Kallio et al., 2016), as

stated in Chapter 3. For this study, audit trails are easily accessible in a few forms from

the start of this research study to the development and reporting of the findings as needed

to show transparency. Part of the audit trail used in this study was the a priori coding

system (Appendix C) protocol used by the researcher for the interview questions and

audio transcription as another means of cross-checking interview data of the participants

with NVivo software. Finally, the handwritten reflexivity notes are accessible as well as

the recordings of the participant interviews for review as needed (Silverman, 2016). I

followed the specific stage by stage process from general note taking during participant

interviews through reflexivity, use of the priori coding in the interview questions that

were used during the data collection process through deductive coding and cross checking

with the NVivo software (Feng & Behar-Horenstein, 2019).

Confirmability

As stated in Chapter 3, confirmability refers to objectivity or the ability of others

to confirm findings (Stahl & King, 2020). Confirmability also refers to the researcher’s

transparency and documentation of processes (Korstjen & Moser, 2018; Nowell et al.,

2017). I show confirmability in this study when I described the process used for data

collection and analysis as suggested by Meadows (2003). I provided a detailed

description of the sequence for data collection, coding, and analysis to deliver a clear and

well-defined accountability for the process as shown throughout this chapter.

Transparency can be defined as “the degree of detail and disclosure about the

specific steps, decisions, and judgment calls made during a scientific study” (O’Kane et

al., 2021, p 105). I provided transparency to the data collection and analysis process by

discussing the levels of queries used within NVivo 12 software. Abu (2016) stated that

this confirms a level of credibility for the researcher that provides trustworthiness of the

data results when using an analytics software platform. This process was described in

detail in this chapter to provide the transparency of the data analysis development that

yielded the main themes and results of the study as seen in Table 4 and Figures 3, 4, and

Trustworthiness in This Study

Trustworthiness, the central concept by which to judge the quality of interpretive

qualitative research is enhanced by demonstrating that researchers understand their

context and data (credibility), showing consistency and lack of bias in data analysis

(confirmability), providing enough detail for possible replication (dependability), and

allowing for assessment of a study’s outcomes in relation to other contexts is

transferability (Korstjen & Moser, 2018; Lichtman, 2017; O’Kane, et al., 2021; Patton,

2014). Maher et al. (2018) stated that NVivo software maximizes researcher data

interaction in a variety of modalities that ensures the analysis process is rigorous and

productive. They further stated that reflection on an authors' research analysis process,

combined with consultation with the literature, would suggest digital analysis software

packages such as NVivo do not fully scaffold the analysis process but provide excellent

data management and retrieval facilities that support analysis and write-up. Further,

Bonello and Meehan (2019) agreed and stated that the NVivo 12 software platform was

intuitive enough to drive intentional queries on the data while showing the trail of

breadcrumbs for researcher credibility and trustworthy results in qualitative study (p.

490).

Another point in achieving trustworthiness for this study was use of a thematic

coding process. Thematic analysis provides a highly flexible approach that can be

modified for the needs of many studies, providing a rich and detailed, yet complex

account of data (Nowell et al., 2017). Thematic analysis is a particularly good choice for

those researchers early in their career and does not require the detailed theoretical and

technological knowledge of other qualitative approaches, it offers a more accessible form

of analysis (Tobin & Begley, 2004). Thematic analysis provides a decision trail, stated

White et al. (2012), that can be shown and presented via narrative or visual display that

enhances the rigor of study with thematic coding process.

Summary

The research question driving this study was: What are the perceptions and

experiences of higher education administrators of 4-year colleges in the application of

student data targeted for tracking and predicting retention? Three main themes emerged

from the inductive coding process driven by the NVivo 12 software platform. The

documents uploaded into the NVivo software for the coding process were the interview

transcripts with initial manual coding completed and my notes and memos written during

the participant interviews. In deciphering and coding the participants’ experiences and

perceptions, the first theme that emerged is that there is no common data mining

definitions nor language for understanding the process for student variable identification

in retention initiatives at each institution. The third theme that emerged from the

participants’ perceptions and experiences is that there is a manual process due to siloed

institutional data.

This third theme also contained two subthemes. The first subtheme was the

manual process for siloed data this further causes challenges in data result turnaround

time. The process to retrieve raw data and analyze in a timely manner requires many

resources that institutions do not have to be a data driven college. The second subtheme

was retention initiatives can only be data driven at a minimal level because of the manual

processes across an institution. These themes and subthemes will be further deliberated in

Chapter 5 in the final discussion, conclusions, and recommendations regarding the results

of this study.

Chapter 5: Discussion, Conclusions, and Recommendations

The purpose of this basic qualitative study was to explore perceptions and

experiences of administrators of 4-year colleges in terms of their application of student

data for tracking and predicting retention. This study could potentially fill a gap in the

literature regarding tracking and predicting processes needed for college retention

initiatives. I conducted interviews with retention administrators at 4-year U.S. colleges to

obtain their experiences and perceptions regarding student data applications for retention

initiatives at their institutions. I manually coded data before uploading the documents into

NVivo 12 for further coding. I reported findings by discussing main themes that emerged.

The themes that emerged for answering the RQ were, that no common data mining

practices or definitions existed, and that student retention decisions are only partially

driven by data, and siloed data is prominent and problematic. The third theme, siloed

data, had two subthemes that were, manual data retrieval is problematic and this further

creates time-consuming data analysis processes. This chapter includes interpretations of

findings as well as limitations of this study. Before concluding this chapter, I include a

brief discussion of recommendations and implications of this study.

Interpretation of the Findings

Interview data were used to provide answers to the RQ for this study. Themes

aligned with peer-reviewed literature regarding retention initiatives in higher education.

In this section, I present interpretations of findings for this study and describe how it

connects to, confirms, and extends what has been found in existing literature.

Theme 1: No Common Data Mining Practices or Definitions

A need in higher education data practices is for a common set of definitions to

inform a general understanding for administrators of data mining practices was a theme

identified in this study. This theme is supported by findings from previous literature and

research that showed a need for higher education to start with basic knowledge of data

mining processes that should be informed by a common set of definitions and policy to

drive research initiatives. Having a set of common data mining definitions to drive

student variable identification for retention initiatives is integral to supporting results that

confirm data transparency in practice. Institutions frequently begin data mining practices

out of order and have a difficult time with faculty buy-in and gaining trust of staff without

first implementing common agreed upon data mining definitions that inform and confirm

processes for analyzing and presenting results (Chaurasia et al., 2018; Gagliardi et al.,

2018).

Knowledge of basic data mining terminology was not widely known by the

participants in this study. Such as the Gandomi and Haider three V construct (2015) that

speaks to the process of data being broken down into prescriptive, predictive, and

descriptive parts and allows decision-makers to see patterns and trends from past

outcomes. Torrecilla and Romo (2018) said common data mining processes are collected

in very different ways, that can be a manual excel sheet process or via the use of software

systems that calculate data through a filtering and specific search language for analysis.

Kwon et al. (2014) said to maintain quality of data, institution-wide data mining

definitions must be in place in higher education to underpin practices and processes. Lack

of participant knowledge as well as their peer administrators was noted as also preventing

colleges from successfully applying student data in retention initiatives.

Theme 2: Student Retention Decisions are Only Partially Driven by Data

Many participants perceived their institution to be inept in terms of data-driven

decision making and this primes institutional leadership to making knee-jerk reactions

rather than being proactive. New processes should be considered for outcome-based

education including data administrators to take an intentional look at student data within

departmental audits to reveal student success measurements and tracking to fuel

datadriven decision making involving retention. Also, many institutions do not have a

culture of data-based decision making, and therefore identification of student data to

assess learning as well as overall retention is problematic I identifying those students who

retain and why (Chaurasia et al., 2018).

Participants said inconsistency in data tracking and unknown origins of data

affects institutional decision-makers in terms of having enlightened institutional data

tracking policies and processes adopted for student success. Niebel et al. (2019) said data-

informed decision-making practices quickly yield benefits to higher education institution

through increased retention, financial returns, and satisfaction of customers.

Theme 3: Siloed Data

Participants in this study stated that a big problem at their institutions was siloed

data. Findings in this study indicated that a point of frustration was a lack of a common or

centralized system for storing and analyzing data. Matsebula and Mnkandla (2017) said

the absence of an appropriate infrastructure for data mining feeds a culture of separated

and individualized data practices, which leads to data tracking failures to support

decision-making that influences student retention and institutional revenue in many facets

of student success in higher education such in academics and social connectedness. Avella

et al. (2016) said data systems that rely upon siloed data and manual practices for data

analysis have proven to be ineffective for time management and tend to be riddled with

human error in terms of tracking student retention in higher learning institutions.

Williamson (2017) said ethical issues in data storage and student information protection

are an ongoing concern. He continued and stated that retention data should be housed in a

centralized data warehouse, and responsibility for its use and application should be

widely dispersed with proper training and accountability.

Subtheme A: Manual Data Retrieval Processes

Participant data showed frustrations with manual data retrieval processes that

come from siloed data practices. participants perceived their institutions as struggling to

make data-informed decisions, and manual processes for retrieving siloed data was one of

those reasons. Tsai et al. (2015) said management, processing, and application of raw

student data cannot be accomplished using simple Excel sheet formulas that were once

considered traditional. Collection and analysis processes for identification of variables in

retention requires a more sophisticated approach then the use of manual paper processes

and siloed data.

Data results from interviews showed a consistent concern for ethical issues

involving data being kept in spreadsheets and files within each department and program,

which leads to ethical issues involving protection of student data. Lacerenza et al. (2018)

said participation in decision-making to promote healthy institutional climates begins

with collecting unbiased and protected student data. Lacerenza et al. said successful

teams produce desired outcomes with clean and safe data variables using demonstrated

and effective processes. Team development interventions are relevant in terms of

institutional survival, but this is contingent upon adherence to data policy and ethical

practices that protect student data.

Subtheme B: Time Consuming Data Analysis Processes

The second subtheme that emerged from the third theme was the time-consuming

process for data analysis that occurs with siloed data. Participants shared that this is a

primary barrier to successful data informed decision making in siloed data practices.

Many of the interviewees stated that decisions must be made whether there is data or not.

But when it takes days or even weeks to track, collect, analyze, and produce an

aggregated result and infographic(s), there are just too many decisions that need to move

forward. Unfortunately, these decisions are forced to be made as a best educated guess.

The hope is that the data that follows confirms the decision. Participants stated that they

see this type of decision-making being done from an institutional level on down to course

level because data is not readily available.

Higher education administrators who value introducing new methods of thinking

as well as a means of data-informed practice are typically met with barriers (Cope &

Kalantzis, 2016). These barriers are associated with the time-consuming task of

gathering, analyzing, and applying data that is siloed and decentralized. Hadwater et al.

(2018) agreed and stated that the barriers that have contributed to the slow adoption of

data mining efforts are institutional policies that do not support a centralized and time

efficient data mining system. This lack of support in policy and practice ultimately

depletes the efforts needed for student success and institutional revenue (Baer & Norris,

2016).

Limitations of the Study

As the sole researcher of this study, I was responsible for collecting, coding,

analyzing, and interpreting the findings. This can create a limitation of this study in that I

am a partial insider researcher because I have a similar professional role to the

participants who the author interviewed. However, I am removed from the community of

which each participant was a part (Fleming, 2018). I did not intentionally make any

decisions to influence the participant interviews. However, I do have a similar role and

knowledge as the participants and may have inadvertently influenced participant

responses. But as an insider, I was able to speak the jargon and pedagogy with which

participants may be familiar, and this allowed for a comfortability to retrieve honest and

open responses.

I followed Walden University’s Institutional Review Board recommendations and

ethical guidelines and used several methods to mitigate my bias and any influence over

this study (Butler, 2016; Cresswell, 2013; Shufutinsky 2020). I kept detailed notes and

memos during the data collection process to review any potential biases I may have had

regarding the participant interview process and data collection, and this provided a tool

for me to engage in self-reflection. I also used reflexivity to evaluate each interaction with

participants while maintain a professional boundary appropriate as the researcher.

Recommendations

I have three recommendations for further research. The first recommendation is to

conduct a study that specifically addresses the experiences and perceptions of

administrators in higher education that can speak to the data culture of their institution.

This study paralleled the topic, but participant perceptions and experiences were

inconsistent as to the data culture in their institutions. Administrator perceptions of

analytics and data mining tends to contribute to its acceptance and use as a medium for

generating variables for institutional research (Chaurasia et al., 2018). Andrews and

Lemons (2015) stated that for a decision-making culture to exist and to be based upon

data evidence, this practice cannot be at the institutional level alone. The data culture

needs to permeate to staff and faculty who are responsible for the day-to-day services

with students. For example, faculty that continue to make decisions based on personal

experiences rather than data need to be brought into the data culture as they have the most

direct knowledge and experience with students (Andres & Lemons, 2017). A good place

to start in driving efficacious data mining in higher education would be to first gauge if

the culture is ripe for such tasks.

A second recommendation would be to quantitatively conduct a study on the level

and type of siloed data that exists in higher education and what would it take to centralize

these efforts. The data results of this study showed that siloed data is problematic, so

much so, that it was a main theme. This theme of siloed data contained two subthemes

that stated barriers in time management and manual data processes still in place.

Additional study to take a deeper dive into siloed data and the use of manual tracking

systems would benefit higher education best practices (Avella, et al., 2016). This

knowledge is needed so that siloed data practices that have proven to be ineffective in

resolving student retention in higher learning institutions are more widely researched and

provide guidance for resolution (Matsebula & Mnkandla, 2017). Williamson (2017)

acknowledged this and stated that educational data science needs predictive measures for

retention but the level of expertise, manual tracking, and available workforce can be a

barrier. Williamson added that retention data should be housed as a centralized data

warehouse, but the responsibility and application of desiloed data should be widely

dispersed with proper training.

My final recommendation would be to address the ethical concerns related to

student data in higher education. Within the context of this study, participants expressed

concerns for student data being on every faculty desk across campus as a result of siloed

data practices. Although this study did address the barriers to siloed data in higher

education as it relates to retention initiatives, the topic of data ethics was beyond the

scope of the stated problem and purpose. So, it is recommended that further study take

place to address this. The nonexistence of the students’ voices regarding the use of data

mining has presented challenges related to the acceptability of student attributes used for

institutional research (Roberts, et al., 2016). Data analysis in higher education related to

students has been comprised by way of using demographic details, enrollment survey

results, student course assessment, and the academic performance of students, among

others (de Freitas et al., 2015). Ethical perspective is needed in higher education research

to consider student participation in decision making to promote a healthy institutional

climate that serves and benefits all stakeholders.

Implications

By gathering the experiences and perceptions of administrators in this study and

how they see their institution through the student retention lens I learned that this gives a

voice to those struggling with organizational change in data culture. Positive

organizational change occurs when individuals achieve their goals and have influence on

others. In this way, the findings can influence those administrators wanting change in

their institution for data mining practices in student success by improving policy and

practice by beginning practices to desilo data, improving data informed decisions making,

and creating a common language for practices when using student data in retention

initiatives. This methodological change in practice can ultimately improve the goal

achievement at the organizational level as well.

Higher education accreditation bodies and funders require data-informed results to

show increasing progress each year. Organizational change can be influenced by the

change in methods of practice. This study provides understanding of both the barriers and

opportunities to begin successful data mining methods for practice and policy change in

higher education that can further influence organizational change. Implications for

change in methods of practice for data mining would first and foremost be an increase in

student success and institutional revenue by incorporating the participant data into

multilevel data informed decision making. Successful achievement of institutional goals

will increase the viability of an institution to accreditors and the surrounding community.

The key recommendations from this study that can influence social change on

organizational and societal levels in higher education, by having open discussion to

define common processes for data mining that increase trust and transparency in retention

initiatives. This can be a positive motivator for organization change that also meets the

societal expectations for data informed results in higher education.

The findings of this study showed that the Gandomi and Haider Three V

Construct (2015) and Attarran et al. (2018) model of analytics used as the conceptual

framework were not widely understood by the participants. The conceptual framework

was successful in providing a common language and process that are considered basic

and integral to any data mining efforts in business and education (Baker & Siemens,

2014; Sivarajah et al. 2017; Williamson, 2017). However, during data collection it was

revealed that the participants knew little about this common process and language. This

implicates that administrators are either being hired with little or no expertise for what is

expected given their title and job description, or that data mining training efforts are

needed.

Although the participants were all from 4-year colleges, the results can be

transferrable to most higher education institutions that struggle with siloed data pools,

little or no data mining processes, and need to increase data informed decision making.

By reviewing methods of practice at varying levels of an institution, the findings of this

study can inform the readers of main themes for improving performance in data mining

initiatives in retention. One of the salient points from this study is to desilo data from

departments and programs as much as possible to centralize the data mining process and

protect student data. This is both integral to organizational and societal change that will

further influence what an institution brings to the table for student success and how they

are viewed by accreditors.

Conclusion

This study explored the perceptions and experiences of administrators of 4-year

colleges in applying student data for tracking and predicting retention. Three themes

emerged that can inform gaps in practice that have been noted in previous research

literature. These gaps in practice have been barriers to student success and institutional

revenue. The lack of common process and definitions for data mining practices

accompanies a lack of transparency and distrust from staff and faculty. By making known

what the processes are for identifying student data, who is responsible, and defining a

common language for those processes, college administrators can open possibilities for

organizational change and success. Improved understanding brought about through this

study can be a first step in productive data mining practices for student success and

retention initiatives. Centralizing data or assigning data responsibilities in a designated

department can increase data productivity and data-informed decision making for

institutions of higher education. Manual data entry and tracking practices from excel

sheets and google docs are devices of the past. The outcomes of this study show that

higher education administrators want efficiency and intentionality from data driven

decisions. With increasing requirements for higher education institutions to produce

reliable data results, time is of the essence. The time has passed for higher education to

simply do what has always been done.

By giving a voice to those struggling with organizational change in regards to data

culture, this provided an understanding of both the barriers and opportunities to begin

successful data mining methods for practice and policy change. This change will

influence higher education data mining practices organizationally speaking that can

further influence community and social change that meets their expectations for data

informed results in higher education to show student success and institutional stability.

This can be accomplished in three integrated steps. The first is to influence organizational

change through methods of practice. This study provides understanding of both the

barriers and opportunities to begin successful data mining for updating methods for

practice and policy change in higher education that can further influence organizational

change. Implications for change in methods of practice for data mining would first and

foremost be the understanding of common data mining language and practice that would

yield an increase in student success and institutional revenue by incorporating the

participant data into multi-level data informed decision making.

Organizations change follows when positive change occurs in updating methods

of practice when individuals achieve their goals and have influence on others. In this way,

the findings can influence those administrators wanting change in their institution for data

mining practices in student success by improving policy and practice by beginning

practices to desilo data, improving data informed decisions making, and creating a

common language for practices when using student variables in retention initiatives.

Societal change follows when organizational change aligns to higher education

accreditation bodies and funders data-informed requirements show increasing progress

each year. Organizational change can be influenced by the change in methods of practice.

Successful achievement of institutional goals will increase the viability of an institution to

accreditors and the surrounding community. The key recommendations from this study

that can influence social change on organizational and societal levels in higher education,

by having open discussion to define common processes for data mining that increase trust

and transparency in retention initiatives. This can be a positive motivator for organization

change that also meets the societal expectations for data informed results in higher

education. The need to desilo data from departments and programs, as well as drive the

institutional culture with data informed decision makers, will yield a foundation of

common language and practice in data mining and the ability to focus on the success of

both students and institutions.

References

Albalowi, A., & A Alhamed, A. (2017, December 7). Big data and learning analytics in

higher education: Demystifying variety, acquisition, storage, NLP and analytics.

https://doi.org/10.1109/ICBDAA.2017.8284118

Alsharari, N.M., & Alshurideh, M.T. (2020). Student retention in higher education: The

role of creativity, emotional intelligence and learner autonomy. International

87  
 
Journal of Educational Management, Vol. #?(Issue #?), p. ##-##?  
https://doi.org/10.1108/IJEM-12-2019-0421  
Amankwaa, L. (2016). Creating protocols for trustworthiness in qualitative research.  
Journal of Cultural Diversity, 23(3), 121-127.  
https://www.researchgate.net/publication/324804792_CREATING_PROTOCOL 
S_FOR_TRUSTWORTHINESS_IN_QUALITATIVE_RESEARCH  
Andrews, T., & Lemons, P. (2015). It’s personal: Biology instructors prioritize personal 
evidence over empirical evidence in teaching decisions. CBE—Life Sciences  
Education, 14(1), 1-18. https://doi.org/10.1187/cbe.14-05-0084  
Arensdorf, J., & Naylor-Tincknell, J. (2016). Beyond the traditional retention data: A 
qualitative study of the social benefits of living learning communities. Learning  
Communities: Research & Practice, Vol. #?(Issue #?), p. ##-##?.  
https://eric.ed.gov/?id=EJ1112861  
Attaran, M., Stark, J., & Stotler, D. (2018). Opportunities and challenges for big data 
analytics in U.S. higher education: A conceptual model for implementation.  
Industry and Higher Education, 32(3), 169-182. 
https://doi.org/10.1177/0950422218770937  
Avella, J. T., Kebritchi, M., Nunn, S. G., & Kanai, T. (2016). Learning analytics methods, 
benefits, and challenges in higher education: A systematic literature review. 
Online Learning, 20(2), 13-29. Online Learning 20.2 Full-Issue pdf.pdf  
(ed.gov)  
Babchuk, W. A. (2017). Qualitative research: A guide to design and implementation (4th 
ed.). Adult Education Quarterly, 67(1), 71–73.  

88  
 
https://doi.org/10.1177/0741713616671930  
Baer, L. L., & Norris, D. M. (2016). A call to action for student success analytics. 
Planning for Higher Education, 44(4), 1-10. A call to action for student success 
analytics. Planning for Higher Education, 44(4), 1-10.  
DATA2018_PRIMER_DOWNLOAD.pdf (naspa.org)  
Baker, R.S., & Siemens, G. (2014). Educational data mining and learning analytics. In K.  
Sawyer (ed.), Cambridge handbook of the learning sciences (2nd ed., pp. 253- 
274).  Cambridge University Press. Educational Data Mining and Learning  
Analytics: Applications to Constructionist Research | SpringerLink  
Barbour, R. (2014). Introducing qualitative research: A student’s guide. SAGE  
Publications. Introducing Qualitative Research: A Student's Guide - SAGE  
Research Methods (waldenulibrary.org)  
Barratt, M. J., Ferris, J. A., & Lenton, S. (2015). Hidden populations, online purposive 
sampling, and external validity: Taking off the blindfold. Field Methods, 27(1), 3- 
21. https://doi.org/10.1177/1525822X14526838  
Bashir, M., Afzal, M.T., & Azeem, M. (2008). Reliability and validity of qualitative and 
operational research paradigm. Pakistan Journal of Statistics and Operation  
Research, 1(1), 35-45. https://doi.org/10.18187/pjsor.v4i1.59  
Bochner, A. (2018). Unfurling rigor: On continuity and change in qualitative inquiry.  
Qualitative Inquiry, 24(6), 359-368. https://doi.org/10.1177/1077800417727766  
Bomatpalli, T. (2013). Significance of big data and analytics in higher education.  
International Journal of Computer Applications, 68, 21-23.  

89  
 
https://doi.org/10.5120/11648-7142  
Bonello, M. & Meehan, B. (2019). Transparency and coherence in a doctoral study case 
analysis: Reflecting on the use of NVivo within a ‘framework’ approach. The 
Qualitative Report, 24(3), 483-498 
https://www.proquest.com/openview/5bc7bedc8740d7875375b49d80688e5c/1?pq 
-origsite=gscholar&cbl=55152  
Bradley, E. H., An, M.-W., & Fox, E. (2020). Reopening colleges during the Coronavirus 
disease 2019 (COVID-19) pandemic—One size does not fit all. JAMA Network  
Open, 3(7), e2017838-e2017838.  
https://doi.org/10.1001/jamanetworkopen.2020.17838  
Braxton, J. M. (2019). Leaving college: Rethinking the causes and cures of student 
attrition by Vincent Tinto (Review). Journal of College Student Development,  
60(1), 129-134. https://doi.org/10.1353/csd.2019.0012  
Bughin, J. (2016). Big data, big bang? Journal of Big Data, 3(1), 2.  
https://doi.org/10.1186/s40537-015-0014-3  
Busteed, B. (2016). Welcome to the exit era of higher education. Gallup.   
https://news.gallup.com/opinion/gallup/199304/welcome-exit-era-
highereducation.aspx  
Butler, J. L. (2016). Rediscovering Husserl: Perspectives on the epoche and the 
reductions. The Qualitative Report, 21(11), 2033-2043. Rediscovering Husserl:  
Perspectives on the Epoché and the Reductions (nova.edu)  

90  
 
Castillo-Montoya, M. (2016). Preparing for interview research: The interview protocol 
refinement framework. The Qualitative Report, 21(5), 811-831. Retrieved from 
http://nsuworks.nova.edu/tqr/vol21/iss5/2  
Centers for Disease Control and Prevention. (2020). Coronavirus Disease 2019 (COVID- 
19). Prevention & treatment. Centers for Disease Control and Prevention.  
https://www.cdc.gov/coronavirus/2019-ncov/prevent-gettingsick/prevention.html.  
Chaurasia, S. S., & Frieda Rosin, A. (2017). From Big Data to Big Impact: Analytics for 
teaching and learning in higher education. Industrial and Commercial Training,  
49(7/8), 321–328. https://doi.org/10.1108/ICT-10-2016-0069  
Chaurasia, S. S., Kodwani, D., Lachhwani, H., & Ketkar, M. A. (2018). Big data 
academic and learning analytics: Connecting the dots for academic excellence in 
higher education. International Journal of Educational Management, 32(6),  
1099–1117. https://doi.org/10.1108/IJEM-08-2017-0199  
Chaurasia, S. S., & Rosin, A. F. (2017). From Big Data to Big Impact: Analytics for 
teaching and learning in higher education. Industrial and Commercial Training. 
https://doi.org/10.1108/ICT-10-2016-0069  
Chen, H., Chiang, R. H. L., & Storey, V. C. (2012). Business Intelligence and Analytics:  
From Big Data to Big Impact. MIS Quarterly, 36(4), 1165–1188. JSTOR. 
https://doi.org/10.2307/41703503  
Conrad, C. F., & Wilson, R. F. (1985). Academic Program Reviews: Institutional  

91  
 
Approaches, Expectations, and Controversies. ASHE-ERIC Higher Education 
Report No. 5, 1985. Association for the Study of Higher Education, Publications 
Department, Washington, DC. https://eric.ed.gov/?id=ED264806.  
Cope, B., & Kalantzis. M. (2016). Big Data Comes to School: Implications for Learning,  
Assessment, and Research. AERA Open 2(2). 
https://doi.org/10.1177/2332858416641907.  
Creswell, J. W. (2013). Qualitative inquiry and research design: Choosing among five 
approaches. Thousand Oaks, CA: Sage  
Creswell, J. W., & Poth, C. (2016). Qualitative Inquiry and Research Design: Choosing  
Among Five Approaches. SAGE Publications.  
Crotty, M. (1998). The Foundations of Social Research. London: Sage.  
Daniel, B. (2015). Big Data and analytics in higher education: Opportunities and 
challenges. British Journal of Educational Technology, 46(5), 904–920. 
https://doi.org/10.1111/bjet.12230  
Davidson, W. B., Beck, H., & Grisaffe, D. B. (2015). Increasing the institutional 
commitment of college students: Enhanced measurement and test of a 
nomological model. Journal of College Student Retention: Research, Theory & 
Practice 17 (2): 162–85. https://doi.org/10.1177/1521025115578230.  
DeCarlo, M. (2018). 13.2 Qualitative interview techniques. Scientific Inquiry in Social  
Work. (Chap 13). Pressbooks. 
https://scientificinquiryinsocialwork.pressbooks.com/chapter/13-2-
qualitativeinterview-techniques/.  

92  
 
de Freitas, S., Gibson, D., Du Plessis, C., Halloran, P., Williams, E., Ambrose, M.,  
Dunwell, I., & Arnab, S. (2016). Foundations of Dynamic Learning Analytics:  
Using University Student Data to Increase Retention. British Journal of 
Educational Technology, 46(6), 1175–1188. Foundations of dynamic learning 
analytics: Using university student data to increase retention - Freitas - 2015 -  
British Journal of Educational Technology - Wiley Online Library  
Dewberry, C., & Jackson, D. (2018). An application of the theory of planned behavior to 
student retention. Journal of Vocational Behavior 107, 100–110.  
https://doi.org/10.1016/j.jvb.2018.03.005.  
Ekowo, M., & Palmer, I. (2016) The promise and peril of predictive analytics in higher 
education: A landscape analysis. New America.  
https://eric.ed.gov/?id=ED570869.  
Elgendy, N., & Elragal, A. (2016). Big Data Analytics in Support of the Decision Making  
Process. Procedia Computer Science, 100, 1071–1084.  
https://doi.org/10.1016/j.procs.2016.09.251  
Elhassan Ali, F. A., & Klett, F. (2018). Learning Analytics in Higher Education—The  
Path Toward Educational Management Intelligent Systems. 2018 11th  
International Conference on Developments in ESystems Engineering (DeSE),  
244–249. https://doi.org/10.1109/DeSE.2018.00050  
Elnozahy, W. A., El Khayat, G.A., Cheniti-Belcadhi, L., &Said, B. (January 1, 2019).  
Question Answering System to Support University Students’ Orientation,  
Recruitment and Retention. Procedia Computer Science, CENTERIS 2019 - 
International Conference on ENTERprise Information Systems / ProjMAN 2019 -  

93  
 
International Conference on Project MANagement. 164: 56–63.  
https://doi.org/10.1016/j.procs.2019.12.154.  
Etikan, I. (2016). Comparison of Convenience Sampling and Purposive Sampling.  
American Journal of Theoretical and Applied Statistics, 5, 1.  
https://doi.org/10.11648/j.ajtas.20160501.11  
Feng, X., & Behar-Horenstein, L. (2019). Maximizing NVivo utilities to analyze 
openended responses. The Qualitative Report, 24(3), 563-571.  
https://nsuworks.nova.edu/tqr/vol24/iss3/11  
Finlay, L. (2013). Unfolding the phenomenological research process: Iterative stages of 
seeing afresh. Journal of Humanistic Psychology 53(2): 172–201.  
https://doi.org/10.1177/0022167812453877.  
Gagliardi, J., Parnell, A., & Carpenter-Hubin, J. (2018). The analytics revolution in higher 
education. Change: The Magazine of Higher Learning, 50(2), 22–29. 
https://doi.org/10.1080/00091383.2018.1483174  
Gandomi, A., & Haider, M. (2015). Beyond the hype: Big data concepts, methods, and 
analytics. International Journal of Information Management, 35(2), 137–144. 
https://doi.org/10.1016/j.ijinfomgt.2014.10.007  
Ganotice, F. A., King, R. B. (2014). Social influences on students’ academic engagement 
and science achievement. Psychological Studies 59 (1), 30–35.  
https://doi.org/10.1007/s12646-013-0215-9.  

94  
 
García-Ros, R., Pérez-González, F., Cavas-Martínez, F., & Tomás, J. M. (2019). Effects 
of pre-college variables and first-year engineering students’ experiences on 
academic achievement and retention: A structural model. International Journal of  
Technology & Design Education 29 (4), 915–28. 
https://doi.org/10.1007/s10798018-9466-z.  
Glesne, C. (2016). Becoming qualitative researchers: An introduction, 5th Edition. In  
Pearson. Pearson.  
Gordon, R. (2020). A basic qualitative research study: How marginalized students’ ability 
to succeed is impacted during freshman and sophomore years in college.  
[Thesis, Concordia University, St. Paul]. Retrieved from 
https://digitalcommons.csp.edu/cup_commons_grad_edd/183  
Griffiths, A. (2017). Forecasting failure: Assessing risks to quality assurance in higher 
education using machine learning [PhD Thesis]. King’s College London. British 
Library EThOS: Forecasting failure : assessing risks to quality assurance in higher 
education using machine learning (bl.uk)  
Guetterman, T. C. (2017). Descriptions of sampling practices within five approaches to 
qualitative research in education and the health sciences. Forum Qualitative  
Sozialforschung / Forum: Qualitative Social Research, 16(2). 
https://doi.org/10.17169/fqs-16.2.2290  
Hadwer, A. A., Gillis, D., & Rezania, D. (2019). Big data analytics for higher education 
in the cloud era. 2019 IEEE 4th International Conference on Big Data Analytics  
(ICBDA), 203–207. https://doi.org/10.1109/ICBDA.2019.8713257  

95  
 
Harackiewicz, J. M., & Priniski, S. J. (2018). Improving student outcomes in higher 
education: The science of targeted intervention. ANNUAL REVIEW OF  
PSYCHOLOGY, VOL 69, 69, 409–435. https://doi.org/10.1146/annurev-psych- 
122216-011725  
Heflin, H., Shewmaker, J., & Nguyen, J. (2017). Impact of mobile technology on student 
attitudes, engagement, and learning. Computers & Education, 107, 91–99. Impact 
of mobile technology on student attitudes, engagement, and learning -  
ScienceDirect (waldenulibrary.org)  
Heiner, C., Heffernan, N., & Barnes, T. (July, 2007). Educational data mining workshop.  
Supplementary Proceedings of the 13th International Conference of Artificial  
Intelligence in Education. Marina del Rey, CA. USA 10.1.1.66.1441.pdf  
(educationaldatamining.org)  
Henke, N., Puri, A., & Saleh, T. (2020). Accelerating Analytics to Navigate COVID‐19 
and the Next Normal. McKinsey & Company. accelerating-analytics-to-
navigatecovid-19.pdf (mckinsey.com)  
Hora, M. T., Bouwma-Gearhart, J., & Park, H.J. (2017). Data driven decision-making in 
the era of accountability: Fostering faculty data cultures for learning. The Review 
of Higher Education 40 (3), 391–426. https://doi.org/10.1353/rhe.2017.0013.  
Hubler, S. (2020, August 15). As colleges move classes online, families rebel against the 
cost. The New York Times. https://www.nytimes.com/2020/08/15/us/covidcollege-
tuition.html  

96  
 
Huda, M., Maseleno, A., Atmotiyoso, A., Siregar, M., Ahmad, R., Jasmi, K., & Muhamad, 
N. (2018). Big data emerging technology: Insights into innovative environment 
for online learning resources. International Journal of Emerging Technologies in 
Learning (IJET) 13 (1), 23–36. (PDF) Big Data Emerging  
Technology: Insights into Innovative Environment for Online Learning Resources  
(researchgate.net)  
Huselid, M., & Minbaeva, D. (2019). In Press: Sage Handbook of Human Resource  
Management (2nd edition), Wilkinson, A., Bacon, Lepak, L., & Snell, S. (Eds.).  
SAGE Reference - The SAGE Handbook of Human Resource Management  
(sagepub.com)  
Kallio, H., Pietilä, A.-M., Johnson, M., & Kangasniemi, M. (2016). Systematic 
methodological review: Developing a framework for a qualitative semi-structured 
interview guide. Journal of Advanced Nursing, 72(12), 2954–2965.  
https//doi.org/10.1111/jan.13031  
Kivunja, C., & Kuyini, A. B. (2017). Understanding and applying research paradigms in 
educational contexts. International Journal of Higher Education, 6(5), 26. 
https://doi.org/10.5430/ijhe.v6n5p26  
Klimek, J. & Klimek, J.A. (2020). IT and data mining in decision-making in the 
organization. Education management in the culture of late modernity. Procedia 
Computer Science 176: 1990–99. https://doi.org/10.1016/j.procs.2020.09.235.  
Knapp, M. S. (2017). The practice of designing qualitative research on educational 
leadership: Notes for emerging scholars and practitioner-scholars. Journal of  

97  
 
Research on Leadership Education 12(1): 26–50.  
https://doi.org/10.1177/1942775116647365.  
Korstjens, I., & Moser, A. (2018). Series: Practical guidance to qualitative research. Part  
4: Trustworthiness and publishing. European Journal of General Practice, 24(1),  
120–124. https://doi.org/10.1080/13814788.2017.1375092  
Korstjens, I., & Moser, A., (2017). Series: Practical guidance to qualitative research. Part 
2: Context, research questions and designs. European Journal of General Practice 
23 (1): 274–79. https://doi.org/10.1080/13814788.2017.1375090.  
Krieb, D. (2018). Assessing the impact of reference assistance and library instruction on 
retention and grades using student tracking technology. Evidence Based Library 
and Information Practice 13(2): 2–12. https://doi.org/10.18438/eblip29402.  
Kwon, O., Lee, N., & Shin, B. (2014). Data quality management, data usage experience 
and acquisition intention of big data analytics. International Journal of  
Information Management, 34(3), 387–394.  
https://doi.org/10.1016/j.ijinfomgt.2014.02.002  
Lacerenza, Christina N., Shannon L. Marlow, Scott I. Tannenbaum, and Eduardo Salas. 
(2018). Team development interventions: Evidence-based approaches for 
improving teamwork. AMERICAN PSYCHOLOGIST  73 (4), 517–31. 
https://doi.org/10.1037/amp0000295.  
Lee, G. (2021). Clarifying the complexities of qualitative research: A book review of 
phillip adu's A step-by-step guide to qualitative data coding. The Qualitative  
Report, 26(7), 2168-2170. doi:http://dx.doi.org/10.46743/2160-3715/2021.4947  

98  
 
Levitt, H. M., Motulsky, S. L., Wertz, F. J., Morrow, S. L., & Ponterotto, J. G. (2017). 
Recommendations for designing and reviewing qualitative research in 
psychology: Promoting methodological integrity. Qualitative Psychology, 4(1), 2– 
22. https://doi.org/10.1037/qup0000082  
Lichtman, M. (2017). Judging the research process and product. Qualitative research for 
the social sciences. SAGE Publications. https://www-
doiorg.ezp.waldenulibrary.org/10.4135/9781544  
Lomet, D. B. (2017). The Future of Data Management. Computer, 50(10), 12–13. 
https://doi.org/10.1109/MC.2017.3641630  
Lourens, A., & Bleazard, D. (2016). Applying Predictive Analytics in Identifying  
Students at Risk: A Case Study. South African Journal of Higher Education, 
30(2), 129–142. https://eds-b-
ebscohostcom.ezp.waldenulibrary.org/eds/pdfviewer/pdfviewer?vid=1&sid=c
9db1d31f1b8-43b0-8c03-9056517fe02b%40pdc-v-sessmgr01  
Madill, A., & Sullivan, P. (2018). Mirrors, portraits, and member checking: Managing 
difficult moments of knowledge exchange in the social sciences. Qualitative  
Psychology (5) 3: 321–39. https://doi.org/10.1037/qup0000089.  
Magaldi, D., & Berler, M. (2018). Semi-Structured interviews. Encyclopedia of  
Personality and Individual Differences, edited by Virgil Zeigler-Hill and Todd K.  
Shackelford. (Chaps 1-6). Cham: Springer International Publishing.  
https://doi.org/10.1007/978-3-319-28099-8_857-1.  
Mah, D. (2016). Learning analytics and digital badges: Potential impact on student 
retention in higher education. Technology, Knowledge and Learning, 21(3), 285- 

99  
 
305. doi:http://dx.doi.org.ezp.waldenulibrary.org/10.1007/s10758-016-9286-8  
Maher, C., Hadfield, M., Hutchings, M., & de Eyto, A. (2018). Ensuring Rigor in  
Qualitative Data Analysis: A Design Research Approach to Coding Combining  
NVivo With Traditional Material Methods. International Journal of Qualitative 
Methods, 17(1), 1–13. 
https://doiorg.ezp.waldenulibrary.org/10.1177/1609406918786362  
Mahroeian, H., Daniel, B., & Butson, R. (2017). The perceptions of the meaning and 
value of analytics in New Zealand higher education institutions. International  
Journal of Educational Technology in Higher Education 14(1): 1–17.  
https://doi.org/10.1186/s41239-017-0073-y.  
Manyanga, F., Sithole, A., & Hanson, S. (2017). Comparison of student retention models 
in undergraduate education from the past eight decades. Journal of Applied  
Learning in Higher Education 7, 30–42. EJ1188373.pdf (ed.gov)  
Matsebula, F., and E. Mnkandla. (2017). A big data architecture for learning analytics in 
higher education. IEEE AFRICON: 951–56.  
https://doi.org/10.1109/AFRCON.2017.8095610.  
Maxwell, J. (2008). Designing a qualitative study. Qualitative Research.   
https://doi.org/10.4135/9781483348858.n7.  
Meadows KA. (2003). So you want to do research? 3. An introduction to qualitative 
methods. British Journal of Community Nursing, 8(10), 464–469.  
https://www.researchgate.net/publication/9037262_So_you_want_to_do_research 
_3_An_introduction_to_qualitative_methods  

100  
 
Merriam, S. B. (2015). Qualitative Research: A Guide to Design and Implementation, 4th  
Edition (4 edition). John Wiley & Sons.  
Merriam, S. B., & Tisdell, E. J. (2016). Qualitative Research: A Guide to Design and  
Implementation. John Wiley & Sons.  
Mills, A., Durepos, G., & Wiebe, E. (2010). Encyclopedia of Case Study Research.  
SAGE Publications. https://doi.org/10.4135/9781412957397.  
Murray, A. (2015). Academic libraries and high-impact practices for student retention: 
Library deans’ perspectives. Portal: Libraries and the Academy 15 (3), 471–87. 
https://doi.org/10.1353/pla.2015.0027.  
Nadasen, D., & List, A. (2017). Predicting four-year student success from two-year 
student data. Big Data and Learning Analytics in Higher Education: Current  
Theory and Practice, edited by Ben Kei Daniel. 221–36. Cham: Springer  
International Publishing. https://doi.org/10.1007/978-3-319-06520-5_13.  
Nathan, S., Newman, C., & Lancaster, K. (2018). Qualitative interviewing. Handbook of 
Research Methods in Health Social Sciences, edited by Pranee Liamputtong, 1– 
20. https://doi.org/10.1007/978-981-10-2779-6_77-1.  
National Center for Education Statistics. (2020). Undergraduate retention and graduation 
rates. 
https://nces.ed.gov/programs/coe/indicator_ctr.asp#:~:text=Retention%20rates%2 
0measure%20the%20percentage,a%20specified%20period%20of%20time  
Neelakantan, S. (2019). ‘Data Analytics Can Save Higher Education’, Say Top College 
Bodies. Technology Solutions That Drive Education. Retrieved April 13, 2020, 

101  
 
from https://edtechmagazine.com/higher/article/2019/11/data-analytics-can-
savehigher-education-say-top-college-bodies  
Niebel, T., Rasel, F., & Viete, S. (2019). BIG data – BIG gains? Understanding the link 
between big data analytics and innovation. Economics of Innovation and New  
Technology, 28(3), 296–316. https://doi.org/10.1080/10438599.2018.1493075  
Nimmagadda, S. L., & Rudra, A. (2017). Managing the Embedded Digital Ecosystems 
(EDE) Using Big Data Paradigm. In B. Kei Daniel (Ed.), Big Data and Learning 
Analytics in Higher Education: Current Theory and Practice (pp. 43–72).  
Springer International Publishing. https://doi.org/10.1007/978-3-319-06520-5_5  
Nowell, L. S., Norris, J. M., White, D. E., & Moules, N. J. (2017). Thematic analysis:  
Striving to meet the trustworthiness criteria. International Journal of Qualitative  
Methods. https://doi.org/10.1177/1609406917733847  
O’Kane, P., Smith, A., Lerman, M.P. (2021). Building transparency and trustworthiness 
in inductive research through computer-aided qualitative data analysis software.  
Organizational Research Methods, 24(1), 104–139.  
https://doi.org/10.1177/1094428119865016  
Olaya, D., Vásquez, J., Maldonado, S., Miranda, J., & Verbeke, W. (2020). Uplift 
1modeling for preventing student dropout in higher education. Decision Support 
Systems. 134. https://doi.org/10.1016/j.dss.2020.113320.  
Patton, M. (2015). Qualitative Research & Evaluation Methods. SAGE.  
https://us.sagepub.com/en-us/nam/qualitative-research-
evaluationmethods/book232962.  

102  
 
Perrotta, K. (2020). Getting HIP: A study on the implementation of asynchronous 
discussion boards as a high-impact practice in online udergraduate survey history 
courses. The Journal of Social Studies Research 44 (2), 209–17.  
https://doi.org/10.1016/j.jssr.2020.02.001.  
Premalatha, K. (2019). Course and program outcomes assessment methods in 
outcomebased education: A review.  Journal of Education 199 (3), 111–27.  
https://doi.org/10.1177/0022057419854351.  
Priniski, S. J., Hecht, C. A., & Harackiewicz, J. M. (2018). Making Learning Personally 
Meaningful: A New Framework for Relevance Research. Journal of Experimental 
Education, 86(1), 11–29. 
https://www.researchgate.net/publication/320478264_Making_Learning_Personal 
ly_Meaningful_A_New_Framework_for_Relevance_Research  
Provencher, A. & Kassel, R. (2019). High-impact practices and sophomore retention:  
Examining the effects of selection bias. Journal of College Student Retention:  
Research, Theory & Practice 21 (2), 221–41. 
https://doi.org/10.1177/1521025117697728.  
Roberts, L., Chang, V., & Gibson, D. (2016). Ethical considerations in adopting a 
university and system wide approach to data and learning analytics. Big Data and 
Learning Analytics in Higher Education: Current Theory and Practice, edited by  
Ben Kei Daniel, 89–108. Cham: Springer International Publishing.  
https://doi.org/10.1007/978-3-319-06520-5_7.  
Romero, C., Ventura, S. (2007). Educational data mining: A survey from 1995 to 2005.  

103  
 
Expert Systems with Applications, 33, 135–146. 
https://www.researchgate.net/publication/222815211_Educational_data_mining_ 
A_survey_from_1995_to_2005  
Rubel, A., & Jones, K. (2017). Data Analytics in Higher Education: Key Concerns and  
Open Questions. University of St. Thomas Journal of Law and Public Policy  
(Minnesota), 11, 25. Rubel Jones Big Data in Higher Education (ssrn.com)  
Sass, D. A., Castro-Villarreal, F., Wilkerson, S., Guerra, N., & Sullivan, J. (2018). A  
Structural Model for Predicting Student Retention. The Review of Higher  
Education, 42(1), 103–135. https://doi.org/10.1353/rhe.2018.0035  
Sanders, L. D., Daly, A.P., & Fitzgerald, K. (2016). Predicting retention, understanding 
attrition: A prospective study of foundation year students. Widening Participation 
& Lifelong Learning 18 (2), 50–83. https://doi.org/10.5456/WPLL.18.2.50.  
Saunders, B., Sim, J., Kingstone, T., Baker, S., Waterfield, J., Bartlam, B., Burroughs, H.,  
& Jinks. C. (2017). Saturation in qualitative research: Exploring its 
conceptualization and operationalization. Quality & Quantity 52(4): 1893–1907. 
https://doi.org/10.1007/s11135-017-0574-8.  
Schneider, M., & Preckel, F. (2017). Variables associated with achievement in higher 
education: A systematic review of meta-analyses. Psychological Bulletin, 143(6): 
565–600. doi: 10.1037/bul0000098.  
Seaman, M. (2011). Bloom’s taxonomy: Its evolution, revision, and use in the field of 
education. Curriculum and Teaching Dialogue. 1–2(29).  

104  
 
Shanahan, J. G., & Dai, L. (2015). Large scale distributed data science using apache 
spark. Proceedings of the 21th ACM SIGKDD International Conference on  
Knowledge Discovery and Data Mining, 2323–2324. 
https://www.researchgate.net/publication/297582750_Large_Scale_Distributed_D 
ata_Science_using_Apache_Spark  
Shein, E. (2020). How universities deploy student data. Communications of the ACM.  
Association for Computing Machinery. doi:10.1145/3381910  
Shufutinsky, A. (2020). Employing use of self for transparency, rigor, trustworthiness, 
and credibility in qualitative organizational research methods. Organizational 
Development Review 52 (1), 50-58.  
https://www.researchgate.net/publication/340539936_Employing_Use_of_Self_fo 
r_Transparency_Rigor_Trustworthiness_and_Credibility_in_Qualitative_Organi 
zational_Research_Methods  
Silverman, D. (2016). Qualitative Research. SAGE.  
Singh, D., & Reddy, C. K. (2015). A survey on platforms for big data analytics. Journal 
of Big Data, 2(1) 8. A survey on platforms for big data analytics | Journal of Big  
Data | Full Text (springeropen.com)  
Sivarajah, U., Kamal, M. M., Irani, Z., & Weerakkody, V. (2017). Critical analysis of big 
data challenges and analytical methods. Journal of Business Research 70 (1),  
263–86. https://doi.org/10.1016/j.jbusres.2016.08.001.  

105  
 
Slater, S., Srećko J., Vitomir K., Ryan S. B,, and Dragan G. (2017). Tools for educational 
data mining: A review. Journal of Educational and Behavioral Statistics 42 (1) 
85–106. https://doi.org/10.3102/1076998616666808.  
Soares, L., Steele, P., & Wayt, L. (2016). Evolving higher education business models.  
American Council on Education: Center for Policy Research and Strategy. 
https://www.acenet.edu/Documents/Evolving-Higher-Education-Business- 
Models.pdf  
Spiers, J., Morse, J.M., Olson, K., Mayan, M., & Barrett, M. (2018).  
Reflection/Commentary on a past article: Verification strategies for establishing 
reliability and validity in qualitative research. International Journal of Qualitative  
Methods 17(1). https://doi.org/10.1177/1609406918788237.  
Stage, F. K., & Hossler, D. (2000). Where is the student? Linking student behaviors, 
college choice and college persistence. In J. M. Braxton (Ed.), Reworking the 
departure puzzle (pp. 170–195) Nashville, TN: Vanderbilt University Press.  
https://www.researchgate.net/publication/282861589_Where_Is_the_Student_Lin 
king_Student_Behaviors_College_Choice_and_College_Persistence  
Stahl, N., & King, J. R. (2020). Expanding approaches for research: Understanding and 
using trustworthiness in qualitative research.  Journal of Developmental  
Education 44 (1): 26–28. 
https://www.researchgate.net/publication/337224462_Expanding_Approaches_for 
_Research_Action_Research  

106  
 
Stefanova, K. & Kabakchieva, D. (2017). Educational data mining perspectives within 
university big data environment. International Conference on Engineering, 
Technology and Innovation (ICE/ITMC), Funchal, 2017, pp. 264-270, doi:  
10.1109/ICE.2017.8279898.  
Swafford, S. J. (2017). Factors affecting retention of first-time, full-time freshmen 
students at higher education institutions within the Appalachian college 
association. University of Tennessee at Chattanooga, Tennessee.  
https://www.researchgate.net/publication/346425936_Expanding_Approaches_for 
_Research_Understanding_and_Using_Trustworthiness_in_Qualitative_Research  
Terrón, M., Olivencia, J. J. L., & Caballero, P. D. F. (2010). Big data irruption in 
education.” Irrupción Del Big Data En La Educación., 57(3) 59–90.  
https://doi.org/10.12795/pixelbit.2020.i57.02.  
Tierney, W. G. (2016). An Anthropological Analysis of Student Participation in College.  
The Journal of Higher Education, 63(6), 603–618. 
https://doi.org/10.1080/00221546.1992.11778391  
Tight, M. (2020). Student retention and engagement in higher education.  Journal of  
Further and Higher Education 44 (5), 689–704. 
https://doi.org/10.1080/0309877X.2019.1576860.  
Tinto, V. (2006). Research and practice of student retention: What next? Journal of  
College Student Retention: Research, Theory & Practice 8 (1),  1–19. 
https://doi.org/10.2190/4YNU-4TMB-22DJ-AN4W.  
Tobin, G. A., Begley, C. M. (2004). Methodological rigour within a qualitative 
framework. Journal of Advanced Nursing, 48, 388–396. doi:10.1111/j.1365- 

107  
 
2648.2004.03207.x  
Torrecilla, J. L., & Romo, J. (2018). Data learning from big data. Statistics & Probability  
Letters: The role of Statistics in the era of big data. 136: 15–19.  
https://doi.org/10.1016/j.spl.2018.02.038.  
Tsai, C.-W., Lai, C.-F., Chao, H.-C., & Vasilakos, A. V. (2015). Big data analytics: A 
survey. Journal of Big Data, 2(1), 21. Big data analytics: a survey | Journal of Big  
Data | Full Text (springeropen.com)  
Tukey, J. W. (1977). Exploratory data analysis. Reading, MA: Addison-Wesley  
Wang, Y. (2017). Education policy research in the big data era: Methodological frontiers, 
misconceptions, and challenges. Education Policy Analysis Archives, 25(94).  
http://dx.doi.org/10.14507/epaa.25.3037  
Wayt, L. & La Cost, B. (2016). Transitioning to performance-based state funding:  
Concerns, commitment, and cautious optimism. Faculty Publications in  
Educational Administration. https://digitalcommons.unl.edu/cehsedadfacpub/112.  
Welsh, E. (2002). Dealing with data: Using NVivo in the qualitative data analysis process. 
Forum, Qualitative Social Research, 3.  
http://www.qualitativeresearch.net/fqs/fqs-eng.htm  
White, D. E., Oelke, N. D., Friesen, S. (2012). Management of a large qualitative data set: 
Establishing trustworthiness of the data. International Journal of Qualitative  
Methods, 11, 244–258. doi:10.1177/160940691201100305  

108  
 
Whitford, E. (2020). Colleges walk back fall reopening plans and opt for online-only                
instruction. https://www.insidehighered.com/news/2020/08/12/colleges-walk-               
back-fall-reopening-plans-and-opt-online-only-instruction  
Wilderotter, R. (2020, July 22). How higher ed can use predictive analytics for student 
recruitment and retention. BPLogix. 
https://www.bplogix.com/blog/predictiveanalytics-student-retention  
Williamson, B. (2017). Who owns educational theory? Big data, algorithms and the 
expert power of education data science. E-Learning and Digital Media 14 (3)105– 
22. https://doi.org/10.1177/2042753017731238.  
Xerri, M. J., Radford, K., & Shacklock, K. (2018). Student engagement in academic 
activities: A social support perspective. Higher Education 75 (4), 589–605.  
https://doi.org/10.1007/s10734-017-0162-9   

Students also viewed