Need an expert in spss

Review_the_Module_3_Measurement_and_Structural_Model_Concepts.pptx

Home >Mathematics homework help >Statistics homework help >Need an expert in spss

Model Samples and Research Methods There are an Infinite number of possibilities mathematically BUT NOT NECESSARILY theoretically!!!

Problem to Test Statistic Chain Relationships

	Definition	The research question	The hypothesis	The methodology	The statistical test	The test statistics
The problem	Identifies an area that needs to be corrected or addressed	Is a rephrasing of the problem into a question	Is a relationship within or between variables that if demonstrated addresses the problem	The methodology is the set of procedures that gathers data within the environment that the problem manifests itself
The research question	A statement, that if answered addresses the problem		Is a relationship that if demonstrated, answers the question	The methodology is a set of procedures that gathers data within the environment where the research questions are be answered
The hypothesis	A relationship within or between constructs/variables that if demonstrated in a statistically acceptable manor, answers the question			Is a means to gather data in a manor that allows addressing the hypothesis and accounts for the environment and sample (population) under investigation
The methodology	A procedures used to gather appropriate data to ensure that the statistical test will actually address the hypothesis				The statistical test is only relevant if appropriate data is gathered using a specifies methodology for the environment and for the sample in which the phenomena is present
The statistical test	A procedure using tools/software that is run using appropriate data gathered using a methodology to address the hypothesis					The test statistic is unique to each statistical procedure

Problem through Test Statistics and Control Chain Examples

Problem	The research question	The hypothesis	The methodology	The statistical test	The test statistics	Controls i.e. factors that are NOT theorized to affect the results	Sample Size
There are positive or negative differences between 2 or more groups in an environment (team, company etc..). The differences are a problem.	Are there differences between 2 or more groups in an environment (team, company, etc..)	The two groups that demonstrate XXX in an environment are different.	Gather a enough data (sample size) related to the problem in the environment for each group depending on the number of variables (indicators) and the number of controls.	The means of each group(s) are different (ANOVA or MANOVA) regardless of the controls i.e. the controls do not matter	Difference in means with a p-value(s). The control groups do not show a difference than the main groups	Age, education, years of experience, position in organization, gender, ethnicity to ensure that these ARE NOT the problem	Obtain samples large enough for the statistical tests AND to break up into groups per the controls
Something(s) is(are) affecting something else in positive or negative way within an environment. The Effect(s) is (are) the problem.	Does XXX (or Do XXX, WWW, and NNN) affect YYYY in an environment (team, company, etc..)	XXX(and WWW NNNn) affect(s) YYYY in a(n) environment (team, company, etc..)	Gather enough data (sample size) related to XX (and WWW, NNN) and gather data related to YYY in the environment depending on the number of variables (indicators) and the number of controls.	XXX (and WWW, NNN) predicts YYYyy (Regression singular or multiple or SEM) regardless of the controls i.e. the controls do not matter	beta weight(s) or path coefficients and statistically significant p-value(s) The control groups do not show a difference.	Age, education, years of experience, position in organization, gender, ethnicity to ensure that these ARE NOT the problem	Obtain samples large enough for the statistical tests AND to break up into groups per the controls
Something(s) is(are) interacting with other things that are then affecting something else in positive or negative way within an environment. Effect(s) is (are) the problem.	Does XXX (or Do XXX, WWW, and NNN) interact with MMM, EEE to affect YYYY in an environment (team, company, etc.)	XXX(and WWW NNNn) affect(s) interacts with MMM or EEE YYYY in a(n) environment (team, company, etc.)	Gather data related to XX (and WWW, NNN), MMM, EEE, and gather data related to YYY (y1, y2 etc.) in the environment depending on the number of variables (indicators) and the number of controls.	XXX (and WWW, NNN) predicts via moderation or mediation with YYY (Regression singular or multiple with moderation or mediation or SEM) regardless of the controls i.e. the controls do not matter	beta weight(s) or path coefficients and p-value(s) of the moderation or mediation are GREATER and statistically significant than without the moderation or mediation. The controls are NOT significant	Age, education, years of experience, position in organization, gender, ethnicity to ensure that these ARE NOT the problem	Obtain samples large enough for the statistical tests AND to break up into groups per the controls

Group differences for the same phenomena Tested using ANOVA i.e. difference in means and p-value

Environment/Population

Group 1

Group 2

Phenomena XXX

(aka variable)

Group n

Indicators

Example

Research Question : In an organization, do different age groups (or education level groups) demonstrate different levels of commitment (the phenomena)?

H: Different age groups (or education level groups) demonstrate different levels of commitment (the phenomena).

Phenomena XXX RELATED to Phenomena YYY Tested via Correlation and p-value

Example

Research Question : In an organization, is the time in the organization (a phenomena) related to the level of commitment (a phenomena)?

H1: Time in the organization (a phenomena) is related to the level of commitment (a phenomena)

Environment/Population

Phenomena

XXX

Phenomena

YYY

Pierson's correlation

and p-Value

Indicators

Phenomena XXX affects (or predicts) Phenomena YYY can be done with Regression or SEM

Example

Research Question : In an organization, does level of commitment affect turnover intention?

H1: In an organization, level of commitment negatively affects turnover intention

Environment/Population

Phenomena

XXX

(Independent

Variable – IV)

Phenomena

YYY

(Dependent

Variable - DV)

Beta Weight

Or and p-Value

Path Coefficient

Indicators

H1-

Phenomena's XXX and WWW affect (or predict) Phenomena YYY Can be Accomplished with Regression or SEM

Environment/Population

Beta Weight

Or and p-Value

Path Coefficient

Indicators

Beta Weight

Or and p-Value

Path Coefficient

Example

Research Question : In an organization, do level of commitment and servant leadership affect turnover intention?

H1: Level of commitment negatively affects turnover intention

H2: Servant leadership negatively affects turnover intention

H1-

H2-

Phenomena

XXX

(Independent

Variable – IV)

Phenomena

WWW

(Independent

Variable – IV)

Phenomena

YYY

(Dependent

Variable -DV)

Phenomena's XXX and WWW affects or predicts Phenomena YYY1 and YYY2 Can be Accomplished with Regression or SEM

Environment/Population

Beta Weight

Or and p-Value

Path Coefficient

Indicators

Beta Weight

Or and p-Value

Path Coefficient

Indicators

Example

Research Question: In an organization, do level of commitment and servant leadership predict turnover intention and productivity?

H1: Level of commitment negatively affects turnover intention

H2: Level of commitment positively affects productivity

H3: Servant leadership negatively affects turnover intention

H4: Servant leadership positively affects productivity

H1-

H2+

H4+

H3-

Phenomena

XXX

(Independent

Variable – IV)

Phenomena

WWW

(Independent

Variable – IV)

Phenomena

YYY1

(Dependent

Variable -DV)

Phenomena

YYY2

(Dependent

Variable -DV)

Phenomena's XXX affects on Phenomena YYY is fully or partially Mediated by Phenomena EEE Can be Accomplished with Regression or SEM

Environment/Population

Beta Weights

Or and p-Values

Path Coefficients

Indicators

Phenomena

EEE

(Mediator)

Indicators

Example

Research Question: In an organization, does Level of Commitment (ME), mediate the relationship between Servant Leadership (IV) and Productivity (DV)?

H1: Servant Leadership positively affects Level of Commitment

H2: Level of Commitment positively affects to productivity

H3: Servant Leadership positively affects to Productivity

Full Mediation: The H3 relationship is zero in presence of Commitment

Partial Mediation: H2 is stronger than H3

H1+

H2+

H3+

Phenomena

XXX

(Independent

Variable – IV)

Phenomena

YYY

(Dependent

Variable - DV)

Phenomena's XXX affects on Phenomena YYY is moderated by Phenomena MMM Can be Accomplished with Regression or SEM

Environment/Population

Beta Weight

Or and p-Value

Path Coefficient

Indicators

Phenomena

MMM

(Moderator)

Indicators

Example

Q: In an organization, does amount of training, moderate the relationship between servant leadership (IV) predict productivity (DV)

H1: Servant Leadership positively affects Productivity

H2: Level of Training increases relationship between Servant Leadership and Productivity

H1+

H2+

Phenomena

XXX

(Independent

Variable – IV)

Phenomena

YYY

(Dependent

Variable - DV)

More Complicated Models Are better suited for SEM Indicators left out for simplicity

Environment/Population

Phenomena

XXX

Phenomena

YYY

Phenomena

EEE1

Phenomena

EEE2

Phenomena

WWW

Phenomena

MMM

Independent

Variable (IV)

Dependent

Variable (DV)

Moderating

Variable (MV)

Independent

Variable (IV)

Mediator

Variable (ME)

Mediator

Variable (ME)

Psychometric Variables vs. Control or Directly Measurable Variables

Psychometric Variable

Measure an individuals “perception” of a Construct

Leadership

Perceived Usefulness

Must be assessed for Reliability and Validity because it is a perception!

Contain measurement error!!

Control Variable/Directly Measured Variables

Are not approximations (i.e. contain no measurement error – they are absolute – always or at specific point in time

Gender

Age

Education

Years in Workforce

Sales for the year

ARE NOT assessed for reliability or validity because they are irrefutable!

contain no measurement error – they are absolute – always or at specific point in time

Not understanding the stages (steps) required for data analysis and why each is accomplished

Data cleansing first – remove bad data or add missing data

Data validity next – ensure we are correctly measuring the variables in our model

Variable relationships last – do the hypothesis hold valid?

Not understanding which tools can be for specific steps

Frequent Student Challenges

Why Do We…..

Clean the data first

To remove data that could produce bad indicators

i.e. some subjects did not answer the questions in a manor consistent with the way other subject did

Add missing data i.e. only one or two indicators from a responder

So that we can make use of a otherwise good data

Assess the indicators and variables next

Because if we are not measuring what we think we are measuring, our hypothesis testing will not be justified!

Assess the variable relationships last

This is the goal!!!!!

Terms that refer to the same ideas or are related

Scale = measures = set of Items = questions related to a variable = questions that measure a construct

This leads to terms and related quality assessment mechanisms such as 1) measurement instrument, or 2) item reliability, 3) scale reliability, or) 4) internal consistency reliability

Construct = factor = variable

The term “Construct” is used in theoretical model – it is a theoretical phenomena of interest that cannot be readily measured

The term “Factor” is used when looking all phenomena that are present in an environment

The term “Variable” refers to the operationalization version of a construct which is measured using indicators

Note by equal ( = ), I mean “related” in some cases

http://www.slideshare.net/DrAkterCMC/reliability-validity-generalizability-and-the-use-of-multiitem-scales

The Operationalized Model – All Reflective Indicators

OBSE

ORG_CARE

FAIRNESS

AUTHORITY

REPUTATION

OrgCare 1

OrgCare 2

OrgCare 3

OrgCare 4

OrgCare 5

Author 1

Author 2

Author 3

Author 4

Author 5

Fair 1

Fair 2

Fair 3

Fair 4

Fair 5

OBSE 1

OBSE 2

OBSE 3

OBSE 4

OBSE 5

REP 1

REP 2

REP 3

REP 4

REP 5

The Measurement Instrument/Measurement Model

(What is Indicator Reliability and Internal Consistency Reliability – What are the Differences?)

OBSE

ORG_CARE

FAIRNESS

AUTHORITY

OrgCare 1

OrgCare 2

OrgCare 3

OrgCare 4

OrgCare 5

Author 1

Author 2

Author 3

Author 4

Author 5

Fair 1

Fair 2

Fair 3

Fair 4

Fair 5

OBSE 1

OBSE 2

OBSE 3

OBSE 4

OBSE 5

REP 1

REP 2

REP 3

REP 4

REP 5

Indicator Reliability -

How well a single indicator measures the variable

Internal Consistency Reliability

A total measure of how well ALL the indictors support measuring the construct/variable

REPUTATION

The Measurement Instrument/Measurement Model

(What is Discriminative Validity?)

Discriminative Validity

Indicators of respective variables correlate most (i.e. associate themselves most) with other indicators from that variable

correlations strongest

and significant

all other individual correlations lower and may be significant or not

The Measurement Instrument/Measurement Model

(What is Convergent Validity?)

Factor 4

Factor 1

Factor 2

Factor 5

OrgCare 1

OrgCare 2

OrgCare 3

OrgCare 4

OrgCare 5

Author 1

Author 2

Author 3

Author 4

Author 5

Fair 1

Fair 2

Fair 3

Fair 4

Fair 5

OBSE 1

OBSE 2

OBSE 3

OBSE 4

OBSE 5

REP 1

REP 2

REP 3

REP 4

REP 5

Factor 3

Convergent Validity

Associated Indicators converge (i.e. align themselves most) on an unnamed factor.

The Regression, Path, or Structural Model

Coefficients, p-values, R2

OBSE

ORG_CARE

FAIRNESS

AUTHORITY

REPUTATION

OrgCare 1

OrgCare 2

OrgCare 3

OrgCare 4

OrgCare 5

Author 1

Author 2

Author 3

Author 4

Author 5

Fair 1

Fair 2

Fair 3

Fair 4

Fair 5

OBSE 1

OBSE 2

OBSE 3

OBSE 4

OBSE 5

REP 1

REP 2

REP 3

REP 4

REP 5

Regression or Beta Weight or

Path Coefficient and p-value

How much the variation in a dependent or mediating variable is explained by the related independent variables.

Regression/Beta Weigh/Path Coefficient

How much change does one variable create on another

and p-value – ex. 0.05

Probability that the result IS by chance

1 minus p-value - ex 0.95

Probability that result holds true

Removal of Outliers – i.e. removal of specific responses not behaving similar to rest

Google

Data Analysis Outliers

What is an Outlier

Two Variable Example

Correlation Plot Tool

Outlier

4 Indicators Example - Box Plot Tool – Sample Size 442

Data Cleansing – Outlier Removal by Variable

i.e. Assess Indicators Associated with One Variable

Outliers

(Response #’)

Indicator # 1

Indicator # 2

Any Variable

Indicator 1

Indicator 2

Indicator 3

Indicator 4

Indicator # 1

Indicator # 2

What kinds of issues have you seen people experiencing?

Area to Assess	Simple Definition	SPSS	Smart PLS
Reliability
Indicator Reliability	Does the indicator consistently measure the variable?	Indicator Loading on Variable Greater than 0.7 and p value > 0.05	Indicator Loading on the Variable Greater than 0.7 and p value > 0.05
Internal Consistency Reliability	Are the items measuring the same thing?	Cronbach Alpha > 0.7	Cronbach Alpha, Composite Reliability, or Average Variance Extracted (AVE) All greater than 0.7
Validity
Convergent Validity	Does a set of indicators represents the same underlying construct	Factor Analysis - Correlations between associated indicators are greater than with others.	Average Variance Extracted (AVE) is greater than 0.5
Discriminant Validity	Do indicators for different constructs only align to their own constructs.	Factor Analysis – Related Indicators load highest on the same factor	AVE of each variable should be greater than squared correlations or Indicators should not have higher correlations with other variables.

Measurement Model Validity - Are we Actually Measuring the Variables?

Reflective Indicators Variables ONLY

Measurement Model Validity - Formative Variables ONLY

Can ONLY be done with SEM Tools - Will NOT work with SPSS

What kinds of issues have you seen people experiencing?

Area to Assess	Simple Definition	SPSS	Smart PLS
Explained Variance (Explanatory Power) (Measured at the affected Variable)	How much the variation in a dependent is explained by the related independent variables.	R2 Values of 0.67, 0.33, and 0.19 as large, moderate, and low levels	R2 Values of 0.67, 0.33, and 0.19 as large, moderate, and low levels
The following are measured on the path between 2 variables
Effect Size	What is the strength of the relationship between 2 variables	Must be Calculated Manually Effect Size = f2= R2/(1-R2) Only simple to do for single (one to one) dependent to Independent Variable relationships Values of 0.35, 0.15, and 0.02 represent large, medium, and small effects	Effect Size = f2 Values of 0.35, 0.15, and 0.02 represent large, medium, and small effects
Predictive Behavior Path Coefficient or Beta Weight	How much change does one variable create on another	Beta Weight with p value Predictive validity variable, we would hope for a beta coefficient coefficient of at least 0.2, p-value less than 0.1, 0.05, 0.01	Path Coefficient with p-value Predictive validity variable, we would hope for a path coefficient of at least 0.2, p-value less than 0.1, 0.05, 0.01
Predictive Behavior Q2	Is the change in the affect variable actually due to changes affecting variable	Not calculated	Predictive Relevance Q2 Should be > 0 and Positive

Structural Model Assessment

Relationships Between Variables

Make sure you understand:

Which tool is used for which data analysis

Which statistic is used to report each data analysis

As you come across articles related to variables you are studying …..pay attention to:

The tools used and

The means to display results

Remember – Not all hypothesis are quantitative – some are qualitative – i.e. demonstrated via rigorous presentation of an argument/defense

Ask Questions throughout this and every class.

Advice

Email Questions to

[email protected]