1
CHAPTER 1: Introduction
Lean management has been a proven management tool for very successful companies
(Ransom, 2007; Kull, Yan, Liu, & Wacker, 2014; Blader, Gartenberg, Henderson, & Prat, 2015),
e.g., Toyota, Totai, or Lundbeck (Houborg, 2010). The central principles of lean management are
a) respect for people and b) continuous improvement (Poksinska, Swartling, & Drotz, 2013;
Houborg, 2010; Liker, & Hoseus, 2010). A supportive organizational culture needs to be created
to establish these principles in an organization (Hatch & Cunliffe, 2006). The creation of such a
“lean” culture depends on the leadership (Mann, 2009). All management levels need to be
engaged during a lean implementation (Ohno, 1982; Emiliani, Stec, Grasso, & Stodder, 2003).
Leadership is important to achieve successful companies, as has been empirically proven (Ohno,
1982; Emiliani et al., 2003). Success is assumed when companies are sustainable and has content
constituents.
A typical implementation problem of lean management is the senior management’s lack
of skill to implement lean (Poksinska et al., 2013; Bashin, 2012a). Bashin (2012a) surveyed 68
manufacturing companies trying to implement lean. He reported, depending on size, that 55% to
66% of the companies have experienced this barrier. Other barriers reported were insufficient
supervisory skills to implement lean (63% to 73%) and employees’ attitude/resistance to change
(60% to 61%). These results supported an earlier survey reporting people-related implementation
problems (Deloitte & Touche, 2002).
In contrast, most companies implementing a lean approach reported having the support of
the top management (e.g., Davis, 2009; Bhuiyan, Baghel, & Wilson, 2006; Roth, 2006.
2
Top management’s support to introduce lean seems to exist, but leaders appear to lack the
ability to introduce/sustain lean (Deloitte & Touche, 2002). Additionally, a cultural-change
challenge seems to exist (e.g. Deloitte & Touche, 2002).
A possible approach to overcome these difficulties might be in using a leadership style,
which supports the two principles of lean. These two principles are respect for people and
continuous improvement (Minter, 2015; Kahlen & Patel, 2011; Li, 2008; Zarbo, 2012; Höök &
Stehn, 2008a; Höök & Stehn, 2008b; Grunden, 2009; Liker & Hoseus, 2010).
Hines, Holweg, and Rich (2004) commented that lean is specifically criticized for its lack
of human integration. Leadership which supports respect for people before initiating continuous
improvement might have better implementation results (MacDuffie, 1995; Sloan, 2011; Dalal,
2010; Hines, 2010; Black, 2009).
Leadership supporting only lean tools achieves low results (Roth, 2006; Aeppel, 2011;
Carter et al., 2011) while implementing lean. Lean tools are typical problem-solving tools like
A3-method, Kanban, Poka-Yoke, and Andon. The lean tools are directed towards the second
principle, continuous improvement. The soft components of continuous improvement, innovative
behavior and proactivity of the employees, seem to be often ignored (e.g. Deloitte &Touche,
2002).
This quandary led to this dissertation on leadership style and its connection to respect and
continuous improvement expressed through proactivity and innovative work behavior.
3
Background
First, a very brief overview is given about the lean management history and lean
leadership. Then the problem statement and purpose of this study follow.
The origin of lean dates to the scientific management approach. Taylor (1911) stated it is
a fallacy to believe that a more efficient production results in more employees out of work, and
he suggested that a management system might provoke employees to work inefficiently to
protect their interests. These statements express the fundamental assumptions of lean
management. The goal of lean management is to increase market share (Atkinson, 2013). Head-
count reduction is not the goal of continuous improvement (Atkinson, 2010). A focus on lean
tools (e.g. Deloitte & Touche, 2002) suppresses the human factor in an organization. The
fundamental issue of scientific management, the neglect of the employees (Littler, 1979), is also
the issue of introducing lean.
Taichi Ohno and Eiji Toyoda created the Toyota Production System, which is recognized
as a lean system of production (Womack, Jones, & Ross, 1990) based on Taylorism. The book
“The machine that changed the world” by Womack, Ross, and Jones (1990) made the Toyota
Production System famous. Imai (1986) popularized KAIZEN (Japanese for “change for the
best”) worldwide, “The starting point for improvement is to recognize the need. This comes from
recognition of a problem. If no problem is recognized, there is no recognition of the need for
improvement. Complacency is the archenemy of KAIZEN.” (p. 9). Globalization drives the
importance of improvement programs and the need for maintaining a competitive advantage
(Sim & Rogers, 2009). Both Imai and Ohno described well what wasteful activities are and how
4
a systematic approach might exist to reduce waste. The waste reduction is the hard-continuous
improvement aspect of lean.
But recently authors like Liker and Convis (2011), Emiliani and Emiliani (2013) and
Dombrowski and Mielke (2013) discussed lean leadership, and the respect for people, an aspect
of lean, and the human aspects of continuous improvement. The principle of respect for people
and the human aspects of the continuous improvement principle seem to be important to
implement and sustain lean because the main identified barriers to lean implementation are
connected to both principles (Bashin, 2012a).
Problem Statement and Statement of Purpose
The leadership theory published by Bass and Bass (2008) described a continuum with
three major leadership styles – transformational leadership, transactional leadership, and
avoidant leadership. A leader is positioned on this continuum and they suggested that the more
transformational a leader is the better. Leadership style might be connected to lean, specifically
the aspects of respect, innovative work behavior, and proactivity of employees. The lean
leadership literature (i.e., Dombrowski, & Mielke, 2013; Dombrowski, & Mielke, 2014; Liker, &
Convis, 2011) does not explain how to measure a lean leadership style or how much it supports
creating sustainable companies with content constituents. One challenge is, to improve the
nebulous knowledge about the relationship of leadership styles developed within the leadership
theory and the principles of lean, mainly developed by practitioners. The fundamental challenge
of this study is to analyze the human aspects of lean.
The significance of the study is supported by following possible implications:
5
For the practitioner a novel way of implementing or sustaining lean might be detected.
For the theory of leadership, a novel approach to describe lean leadership might be developed.
Also, within the leadership theory, it might be detected that different styles than currently
proposed should be combined.
The current analytical research study will shed light on the relationship between
leadership style and lean, specifically the two principals, a) respect for people expressed through
respect, and b) continuous improvement expressed through innovative work behavior and
proactivity.
6
CHAPTER 2: Review of Literature
The scholarship on lean leadership is rather fragmented. A lot of articles have been
published on lean leadership, but an overview of the existing literature does not exist. Therefore,
this review of literature aimed to develop a lean leadership model by conducting a systematic
literature review. The conducted research based on this leadership model by connecting four
variables to the model. Therefore, this literature review provides further a brief overview of a
lean culture as organizational culture and an in-depth review of the literature on the used
research variables, respect, proactivity, innovative work behavior, and leadership style. Finally,
the research questions are presented.
Systematic Literature Review Model
Methods how to conduct systematic literature reviews are presented here. Based on the
existing review models, a method was chosen.
Crossan and Apaydin (2010) used a three-step process to conduct a systematic review of
literature in the field of management. They first did data collection using a pre-defined
algorithm. Next, they analyzed the data using a descriptive statistical method sacrificing depth
for breadth by using pattern-matching and explanation building. Lastly, they synthesized the
data– producing new knowledge as a product of step 1 and 2.
This review of literature followed the approach of Crossan and Apaydin (2010) because it
aligned well both in the context of the body of the sample data to be analyzed and because the
literature lacked the statistical documentation found in software and medical literature.
7
Step 1: Literature Review Data Collection.
The research study used the three-step algorithm process as outlined in Crossan and
Apaydin (2010) to conduct a systematic review by first predefining a selection algorithm to
reduce the subjectivity of the data collection. A Library Assistant Professor at the University of
New Mexico helped to develop search algorithms by using defined strings. Following search
terms were defined: “lean leadership”, “lean management” AND “leadership”, and “lean
management” AND “culture”. The first two search strings are terms used in defining lean
leadership. Based on a definition of leadership by Schein (2010) where he states that leadership
defines the culture of a company, the third search string “Culture” was introduced. The search
terms produced enough hits on google scholar that a systematic literature seemed justifiable.
Lean leadership produced about 1,470 results, “lean management” AND “leadership” about
10,800 results, and “lean management” AND “culture” about 10,900 results (see Figure 1).
8
Lean leadership
articles
1470
20 articles
Peer
reviewed
Lean
management
AND leadership
articles
Around 10800
131 articles
Peer
reviewed
No
repeats
122 articles
Lean
management
AND
culture
articles
Around 10900
162 articles
Peer
reviewed
No
repeats
132 articles
No
repeats
20 articles
Total
274 articles
+
+
=
Figure 1: Systematic Literature Review - Search
The search was reduced to peer-reviewed studies on two databases: Web of Science, and
Business Source Complete. This reduction led to 313 articles in total. Then all the same articles
were eliminated, leading to a total of 274 articles, see figure 1. From the 274 articles, six articles
could not be retrieved from the publisher. The library of the University of New Mexico tried to
receive a library subscription, but the publisher only offered personal subscriptions. The articles
were excluded as the publisher seemed to be only acting as a business improvement magazine
and did not appear to be academically engaged. Five articles were excluded from a title
examination, e.g., a medical content was presented about reducing weight (lean management), or
9
they were marketing materials to sell training about lean management. During the literature
review process, 35 articles were excluded because they did not inform the phenomenon of lean
leadership but used the concept of lean leadership to explain other ideas. Another exclusion
criterion was that the articles needed to be in Spanish, English, or German. One article did not
fulfill this criterion (see figure 2).
Total
274 articles
Relevant titles
retrieved
268
Relevant abstracts
retrieved
263
Articles included
227
Articles not retrieved
from publisher (n = 6)
Exclusion after title
examination (n = 5)
Exclusion after paper
examination (n = 36)
Figure 2: Systematic Literature Review - Exclusion
10
In Figure 3, the publishing schedule of the remaining articles can be seen.
Figure 3: Reviewed Article Timeline
Figure 3 shows that until 2006, there were not very many articles published about lean
leadership, but since then, the numbers of articles increased substantially. This pattern is an
indicator that the interest in the concept of lean leadership has grown and continues to grow.
In aligning with the Systematic Literature review model, the next step in the review
process will show how the data analysis and data synthesis consolidated the literature body and
developed the proposed lean leadership model.
0
10
20
30
40
50
60
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
11
Step 2: Literature Review Data Analysis.
The analysis of this review was based on pattern matching and explanation building
(Yin, 1994). Following Yin (1994) pattern matching is not a precise science and he suggested the
researcher should detect gross matches or mismatches in which an “eyeballing” technique is
sufficient to support conclusions. All selected articles were analyzed to find an overall concept of
lean leadership and its corresponding constructs. The articles were clustered into several types of
articles (see Figure 4).
Figure 4: Types of Articles
The clusters of articles were analyzed in the following sequence: quantitative, surveys,
qualitative, case studies, meta-analysis, interviews, practitioner reports, lit review, theory. The
reason was to identify the lean leadership constructs used in the quantitative, qualitative and
survey studies given that these are easier to identify in this kind of articles. Then use this list of
potential lean leadership constructs to find support for them in the rest of the cluster of articles
except for the theory articles.
53
20 16
96
23
1
11 61
0
20
40
60
80
100
120
12
The theory articles were taken last to verify and expand the developed conceptual lean
leadership model. But they did not provide evidence for the theories presented.
Step 3: Literature Review Data Synthesis.
The new conceptual knowledge produced is the primary value-added product of this
review. During the data integration process, it was evident that a quantitative approach was not
feasible as in medical studies. Medical systematic literature reviews use mainly correlation data,
which can be aggregated. The articles of this systematic literature review did not have enough of
this kind of data. Thus, a qualitative approach was taken (Clark, & Creswell, 2010). First nodes
were identified by analyzing qualitative, survey, and qualitative articles. Based on the nodes, all
literature was analyzed and if a node was supported by at least two articles with more than three
authors in total, a finding was assumed. Based on the findings a lean leadership model was
created (see Figure 5). This model supported the selection of research variables outlined in
Figure 5.
13
External Factors
• National Culture Collectivism vs. Individualism
• Market forces
Organizational Outcomes
Lowering costs Improving quality
Faster delivery More reliable delivery
More flexible product mix High employee engagement
Financial success
Leadership
Beliefs
• Continuous improvement is good
• Importance of quality
• Training of people is valuable
• Lean management is a system
• Importance of customers
• People have positive impact on organization
• Collaboration is good
• Coaching of employees is good
• Experimenting is good
• A growth strategy for the organization is good
• Striving for perfection is good
Organizational Factors
Warranty of job security Trust between management and employees
Transformational Leadership Learning of Kaizen
Training of employees for job Training of employees about problem solving (lean tools)
Training of all leaders about lean management and lean tools Development of leaders
Self-development of leaders through practice No-blame culture
Value stream thinking Hoshin Kanri (all goals are aligned on all levels)
Measurement of understandable non-financial goals Problem solving goals are simplifying and waste reduction
Problem solving methodology in place All organizational levels engaged in continuous improvement
(Kaizen)
Employees participate in problem solving Standardization
Job rotation Gemba
Communication – management listens, employees talk Shop floor management
Visual assistance Importance of front-line supervisor
Workplace organization Leadership development
Cell production / one-piece flow Pull production planning
Value stream accounting (not Overhead accounting) Product development front loaded
NO formal suggestion system
Leadership
Behaviors
• Listening, sharing, open communication
• Using processes to improve organization
• Coaching new behaviors
• Making oneself replaceable
• Being visible to all employees (Gemba)
• Celebrating teams not managers
• Asking why not who (no blame culture)
• Being a role model in modesty, engagement,
respect, and support of staff
• Participating in self-development
• Thinking in value streams
• Promoting one-piece flow and pull principle
• Being a modest leader
• Showing a common vision
• Living integrity
Leadership
Values
• Respect to people
• Trust in employees
• Standardized habits
• Honesty
• Fairness
• Commitment
• Intellectual curiosity
• Safety of employees
Figure 5: Lean Leadership Model
14
Culture of Lean
The organizational culture of lean is not well defined. The focus of lean culture seems to
be on people and processes (Minter, 2015; Kahlen & Patel, 2011; Li, 2008; Zarbo, 2012; Höök &
Stehn, 2008a; Grunden, 2009). Liker and Hoseus (2010) deviated slightly by stating that respect
for people (Houborg, 2010; Raines, 2011) and Kaizen were the central pillars of a lean culture
(Sturdevant, 2014). The differences between process and Kaizen emphasis might be based on a
difference in opinion what the continuous improvement aspect is. Kaizen is Japanese for
continuous improvement; most continuous improvement is based on process thinking. This study
will follow the thought of Kaizen as it seems to be the more inclusive term.
The key aspect, people in an organization willing to improve productivity, was supported
by the research of MacDuffie (1995) and others (e.g., Sloan, 2011; Dalal, 2010; Hines, 2010;
Black, 2009). MacDuffie (1995) had not found an answer to the claim of “pressing out”
employees (Jones, Latham, & Betta, 2013; Corbett, 2013) in lean nor to the counterclaim of
“working smarter.” But he supported Arthur’s (1992) thought of fit between the human
resourceaspects and production strategy.
The central message is lean is not a tool nor a cost reduction strategy (Liker, & Franz,
2012; Atkinson, 2013; Markovitz, 2016), but a management system rooted in respect for people
and search for continuous improvements (Liker & Franz, 2012; Clark, 2016).
Respect seems to be the most important concept in the lean culture. The continuous
improvement concept will be analyzed through two lenses, proactivity, and innovative work
behavior. The assumption is that employees need to be proactive for being able to conduct
innovation.
15
Respect
Respect might have multiple facets. What is the object of respect? How is respect shown?
Authors tried to described respect, and the two central aspects of respect seem to be attitude and
behavior.
Van Quaquebeke and Echloff (2010) defined respect as “a person’s attitude towards other
people, in whom he/she sees a reason that, in itself, justifies a degree of attention and a type of
behavior that in return engenders in the target a feeling of being appreciated in importance and
worth as a person” (p. 344). They emphasized the importance of noticing and understanding
others and connected respect with Kant’s (1788) categorical imperative – an employee is an end
in itself and not a tool to reach a goal. Downie and Telfer (1969) and Frankena (1986) stated,
maintaining a cooperative attitude towards others and acting on it shows respect. Even if the
object of respect is disliked, the cooperative attitude to the object of respect still shows respect
(Simon, & Stürmer, 2003; Spears, Ellemers, & Doosje; 2005). Additionally, Spears, Ellemers,
and Doosje (2005) reasoned that one’s own behavior communicates respect for others.
Respect – Attitude.
Attitudes are artifacts, which are based on values, which are based on beliefs (Schein,
1985a). The following concepts out of in the earlier developed lean leadership model, see figure
5, seem to be connected to the attitude aspect of respect; beliefs, values, and artifacts.
Beliefs.
Within a lean culture, the leaders have to believe in the development of employees
(Sarkar, 2011; Black, 2009; Liker, & Franz, 2012), collaboration (Sarkar, 2011; Liker, & Franz,
2012), and coaching of employees (Sarkar, 2011; Petersen, 2010; Liker, & Franz, 2012).
16
Values.
Hartwell and Roth (2010) reported values of one lean CEO: honesty, fairness,
commitment, respect, and encouragement of intellectual curiosity. The safety of employees and
others is paramount for lean leaders (Liker, & Hoseus, 2010). Modesty seems to be a unique
feature of lean compared to other successful companies (Bortolotti, Boscari, & Danese, 2015).
Lean leaders shall accept that they do not have to solve all problems, but their people (Zarbo,
2012). Additionally, a trusting partnership between management and employees is central and
should be valued by leaders (Patrick, 2015; Mróz, 2010). This aspect might be foundational to
create respect for people.
Artifacts.
Artifacts are everything what an outsider of an organization can easily identify within an
organization (Schein, 1985b). This includes symbols, things, rituals etc.. The following artifacts
support the lean principle of continuous improvement on a personal level.
The commitment to self-development is mandatory for lean leaders by achieving self-
awareness to learn and to improve their knowledge about the Kaizen mind, gemba, teamwork,
and respect for people (Dombrowski, & Mielke, 2013; Dombrowski, & Mielke, 2014, Emiliani,
2008; Liker, & Morgan, 2006; Sturdevant, 2014). Carter (2008) additionally suggested that
management needs to be humble enough to change earlier outcomes.
Respect – Behavior.
Following Schein (1985a) behaviors are artifacts, which should be observable. The
following concepts out of the lean leadership model seem to connect to the behavioral aspects of
respect.
17
Lean leaders value that all employees go home with a sense of fulfillment (Patrick, 2015;
Mróz, 2010). Management creates habits to achieve extraordinary results (Patrick, 2015; Liker,
& Franz, 2012; Teresko, 2008) like admitting shortcomings, delivering commitments,
communicating honestly, sharing ideas, ensuring team member’s understanding, and acting as a
responsible corporate citizen (Liker, & Hoseus, 2010).
In the daily communication, lean leaders need to share and listen (Patrick, 2015; Li,
2008; Gingerich, 2008; Hach, 2009; Hogan, 2009; Raines, 2011; Muthukumar, Tamizhjyothi, &
Nachiappan, 2014; Netland, 2016). A common challenge for leaders is to change from top-down
work order approach to a bottom-up problem reporting communication approach (Glossmann,
Schliebusch, Diehl, & Walshe, 2000; Kahle, 2015).
Lean leaders coach others to develop and apply their talents (Patrick, 2015; Li, 2008;
Kavanagh, & Cole, 2013). The true top-down management activity is coaching and teaching
(Liker, & Franz, 2012; Hach, 2009; Angelis, Conti, Copper, & Gill, 2011; Badurdeen, &
Gregory, 2012; Sturdevant, 2014; Markovitz, 2016) instead of ordering (Clark, 2016; Liker, &
Hoseus, 2010; Blader et al., 2015) or problem-solving for others (Liker, & Franz, 2012; Delisle,
&Turner, 2010; Badurdeen, & Gregory, 2012). Lean leaders shall make themselves replaceable
(Dombrowski, & Mielke, 2014) by coaching others (Flinchbaugh, Carlino, & Curtis-Hendley,
2008; Mann, 2009; Dombrowski, & Mielke, 2013). Bashin (2012a) cautioned that this behavioral
change is difficult for managers.
Jusko (2012) suggested the CEO should be on the floor shop conducting gemba. This
thought is supported by several authors (Robinson, & Kirsch, 2015; Hogan, 2009; Delisle, &
Turner, 2010; Mcloughlin, 2015). Mann (2009) suggested that executive leaders should do two
gemba walks per week for 45 to 60 minutes and a lean teacher shall accompany the leader during
18
the first six to twelve months. But daily practices of lean management seem to be undervalued by
senior managers (Emiliani, 2008; Flinchbaugh et al., 2008) yet should be practiced by the lean
leader (Liker, & Franz, 2012).
Team achievements shall be recognized and celebrated (Patrick, 2015; Li, 2008; Zarbo
2012; Vago, Bell, & Thompson, 2016; Liker, & Morgan, 2006; Li, 2008) and not the manager.
Lean leaders shall show modesty and support Kaizen, but not intervening directly in
problem-solving (Shiba, Graham, & Walden, 1993, Aij, Visse, & Widdershoven, 2015;
Dombrowski, & Mielke, 2014; Poksinska et al., 2013; Kenworthy, 2013). Thus, they coach how
to problem solve, but not problem solve personally. Lean leaders are role models (Mann, 2009;
Dombrowski, & Mielke, 2013; Poksinska et al., 2013, Jusko, 2012; Zarbo, 2012) by showing
modesty, engaging in gemba, respecting staff and supporting staff. They do not solve problems
but teach how to solve problems (Mann, 2009).
Proactivity
Following Atkinson (2013) an active approach is required to building genuine and robust
processes through continuous improvement. For an active approach, leadership needs to allow
employees to be proactive.
Proactive behavior is a construct used in psychology and organizational behavior
literature (e.g., Bowers, 1973, Schneider, 1983, Bandura, 1986). Lewin (1938) stated that all
behavior has both personal and situational causes. Following White (1959) and Langer (1983),
the proactive dimension of behavior is based on one’s need to change and control the
environment. This statement was supported, e.g. by Bandura (1986). Bandura (1986) suggested
action and individuals intentionally drive human activity and humans have the opportunity to
change their surroundings. The importance of proactive behavior for changing an organization
19
has been discussed in the literature (e.g., Miles, Snow, Meyer, & Coleman, 1978). And the
continuous improvement lean principle requires a constantly changing organization.
Bateman and Crant (1993) defined a proactive personality scale as the ability to
intentionally and directly change current circumstances, social or nonsocial. Proactive
individuals might exhibit behaviors of 1. problem finding and idea championing (Maynes,
McCall, & Kaplan, 1985), 2. innovating (Hirschman, 1970), 3. task revision (Staw, & Boettger,
1990), and 4. affect change (Grant, & Ashford, 2008).
The following will show the four aspects of a proactive personality and their relationship
to the lean leadership model.
Problem Finding and Idea Championing.
In the daily communication lean leaders need to share and listen (Patrick, 2015; Li, 2008;
Gingerich, 2008; Hach, 2009; Hogan, 2009; Raines, 2011; Muthukumar et al., 2014; Netland,
2016). A typical challenge for leaders is to change from a top-down work order communication
to a bottom-up problem reporting interface (Glossmann et al., 2000; Kahle, 2015).
Team achievements are recognized and celebrated (Patrick, 2015; Li, 2008; Zarbo 2012;
Vago et al., 2016; Liker, & Morgan, 2006; Lim 2008) and not the manager.
Innovating.
Innovation in lean does not seem to be revolutionary but developmental expressed
through continuous improvement (Mehri, 2006). The presented three aspects are not novel for
lean companies, but non-lean companies might not use them. The improvement of processes is
based on following overarching ideas:
20
Value stream thinking should be one central aspect of the lean leader’s overall
management philosophy and drive innovation (Womack, Jones & Ross, 1990; Mann, 2009;
Novac, & Mihalcea, 2014; Epsten, 2013; Faulkner, 2013; Sarkar, 2011; Atkinson, 2013;
Mcloughlin, 2015; Markovits, 2016). And case studies supported the usage of value-stream
mapping (Thomas, 2016; Harris, & Harris, 2015).
One-piece flow production which may require a cell-production layout is another central
aspect for lean leaders (Minter, 2010; Davidson, 2008; Zarbo, 2012; Davidson, & MacKay,
2009).
The pull principle seems to be the third central aspect of the lean leader’s overall
understanding of process functioning (Womack et al., 1990; Mann 2009, Epsten, 2013; Zarbo,
2012).
Task Revision.
Striving for perfection is essential within lean (Womack, Jones & Ross, 1990;
Flinchbaugh et al., 2008; Dombrowski, & Mielke, 2013; Zarbo, 2012; Liker, & Morgan, 2006;
Henderson, 2008; Gingerich, 2008; Adrian, 2011; Sturdevant, 2014). And this is not negotiable
(Black, 2009; Hogan, 2009). It implies to be open to revising all tasks within an organization.
New approaches to work processes have the risk of failure. So, failure is possible
(Dombrowski, & Mielke, 2013; Dombrowski, & Mielke, 2014; Dawson, 2008). This risk has to
be accepted by managers, and the implementation of new approaches still has to be supported
(Mann, 2009; Flinchbaugh et al., 2008; Simon, & Canacari, 2012; Liker, & Franz, 2012).
Problems are opportunities to learn and improve (Rahn, 2015; Ludwig, 2014; Wyton, & Payne,
2014; Kenworthy, 2013; Liker, & Franz, 2012); employees are part of the solution and not the
problem (Robinson, & Kirsch, 2016; Liker, & Hoseus, 2010; Thomas, 2016; Delisle, & Turner,
21
2010). So, leaders should ask how to improve (Thomas, 2016). Also, going back to a former
solution is viable (Johns, 2015; Carter, 2008).
To Affect Change.
If lean leaders have now accepted that problems exist, improvement is guided by
fundamental process thoughts, and tasks get revised, how can the lean leader make the change
happen?
Lean leaders use cross-functional interactions to improve processes (Patrick, 2015; Zarbo
2012; Li 2008). And the change process is the responsibility of the work groups and not the
management (Liker, & Franz, 2012; Li 2008).
Lean Leaders coach others to develop and apply their talents (Patrick, 2015; Li, 2008;
Kavanagh, & Cole, 2013). The true top-down management activity is coaching and teaching
(Liker, & Franz, 2012; Hach, 2009; Angelis et al., 2011; Badurdeen, & Gregory, 2012;
Sturdevant, 2014; Markovitz, 2016) instead of ordering (Clark, 2016; Liker, & Hoseus, 2010;
Blader et al., 2015) or problem-solving for others (Liker, & Franz, 2012; Delisle, & Turner,
2010; Badurdeen, & Gregory, 2012). Bashin (2012b) cautioned that this behavioral change is
difficult for managers.
Gemba means that a person needs to physically go to the place where an issue might
happen and observe the issue happening (Imai, 1986). Jusko (2012) suggested the CEO should
be on the floor shop conducting gemba. This thought is supported by several authors (Robinson,
& Kirsch, 2015; Hogan, 2009; Delisle, & Turner, 2010; Mcloughlin, 2015).
22
Lean leaders are role models (Mann, 2009; Dombrowski, & Mielke, 2013; Poksinska et
al., 2013, Jusko, 2012; Zarbo, 2012) by showing modesty, engaging in gemba, respecting staff
and supporting staff not in solving problems but learning how to solve problems (Mann, 2009).
The suggested lean leadership model seems to support the four characteristics of a pro-
active personality.
Innovative Work Behavior
Innovating was already mentioned in the proactivity aspect of the lean culture. But
innovative work behavior describes this aspect much deeper.
Innovative Work Behavior is defined through individual employees who “develop, carry,
react to, and modify ideas” (Van de Ven, 1986, p. 592) and create ideas (Scott, & Bruce, 1994).
In the lean organization, this is expressed through Kaizen, Japanese for continuous improvement.
Thus, a lean organization needs to train their employees in Kaizen (Weinstein, 2014; Jusko,
2012; Zarbo, 2012; Hogan, 2009). This kind of behavior should develop employees who can
develop, carry, react to, and modify ideas.
Problem-solving promotes continuous improvement (Lawell, 2010). Visuals information
like andon, metrics, and other lean tools are used on the shop floor to detect problems (Liker, &
Franz, 2012; Thomas, 2016; Liker, & Morgan, 2006; Myszewski, 2015). The goal is to detect
problems or abnormalities from the standard (Robinson, & Kirsch, 2015; Clark, 2016; Liker, &
Franz, 2012; Badurdeen, & Gregory, 2012). Workers shall participate in problems solving. Thus,
not only the process gets improved (Garza-Reyes, Ates, & Kumar, 2015; Robinson, & Kirsch,
2015; Simon, & Canacari, 2012; ; Höök, & Stehn, 2008; Brandt, 2008; Carter, 2008; Hach, 2009;
Hogan, 2009; Peterson, 2010; Allen, 2010; Severs, 2010; Raines, 2011; Smith, 2012; Alves,
23
Dinis-Carvalho, & Sousa, 2012; Kavanagh, & Cole, 2013). But also the people get developed in
their problem-solving abilities (Liker, & Franz, 2012; Henderson, 2008).
The development should take place during gemba (Flinchbaugh et al., 2008;
Dombrowski, & Mielke, 2014; Emiliani, & Emiliani, 2013; Hogan, 2009; Carr, Lawler, & Reny,
2012). And the learning group should have a low leader to employee ratio (Dombrowski, &
Mielke, 2014). The learning event should follow a standardized agenda (Mann, 2009; Poksinska
et al., 2013; Vago et al., 2016; Liker, & Morgan, 2006). Individualized learning in short cycles is
suggested (Dombrowski, & Mielke, 2014; Poksinska et al., 2013; Delisle, & Turner, 2010). And
the learning event should conclude with internal knowledge exchange (Glover, Farris, & Van
Anken, 2015).
The use of PDCA (plan do check act - cycle) is suggested to problem solve
(Dombrowski, & Mielke, 2013; Hillberg, 2015; Ross, 2014; Wyton, & Payne, 2014; Southworth,
2012; Myrvold, 2011; Zarbo, 2012; Vago et al., 2016; Liker, & Morgan, 2006; Howell, 2015).
While using this approach, the root cause analysis is a central element of work during the
planning phase (Rahn, 2015; Liker, & Franz, 2012). Managers may avoid this work to cover up
incidents or to cut corners (Collins, 2010). FMEA (Fehler Moeglichkeiten Einfluss Analyse) is
one methodology to conduct root cause analysis (Rahn, 2015). The C of PDCA is the Gemba
walk, and one should wait and observe until the problem occurs (Southworth, 2012; Liker, &
Franz, 2012; Tatham, 2008). During the A phase, it is suggested to have a lesson learned session
(Glover, Farris, & Van Anken, 2015).
Other methods to problem solve are the A3 method (Faulkner, 2013; Clark, 2016;
Delisle, & Turner, 2010) and the tool six sigma (Miguel, & Carvalho, 2014). Six sigma supports
24
the experimental character of improvement (Spear, 2004; Dawson, 2008; Davidson, & MacKay,
2009).
The management should support the learning of problem-solving through daily activities
(Jusko, 2012; Sarkar, 2011; Myrvold, 2011; Hartwell, & Roth, 2010; Hines, 2010; Delisle, &
Turner, 2010; Kahle, 2015). These activities might be checking the conditions, reviewing the
objectives, and pointing out problems (Mann, 2009; McCreary, 2010; Poksinska et al., 2013;
Rahn, 2015; Kenworthy, 2013; Raines, 2011).
A specific area of learning should be safety (Semiklose, 2014; Raines, 2011; Petersen,
2010; Mróz. 2010; Jusko, 2012).
The success of problem-solving can be measured by developed standards (Liker, &
Morgan, 2006) and is supported by the six-sigma methodology (Bessette, 2012).
The management should take into consideration, what type of Kaizen takes place, Blitz-
Kaizen or developmental Kaizen. It might be preferable to implement fast, small improvements
than waiting for a perfect solution (Delisle, & Turner, 2010; Marksberry, 2012). But Hines
(2010) and Black (2009) cautioned managers, Blitz-Kaizen lead to short-term effects but are not
sustainable. The developmental Kaizen approach seems to be more promising (Liker, & Morgan,
2006; Atkinson, 2013).
A fundamental misunderstanding of Kaizen learning is that it can be achieved through
activities only in a classroom (Liker, & Franz, 2012). The management participating in Kaizen
should ask questions and let the front-line employees problem-solve on the floor shop
(Gingerich, 2008; Netland, 2016). The presence of management is critical during Kaizen
(Cameron-Strother, 2010; Liker, & Hoseus, 2010; Delisle, & Turner, 2010).
25
The continuous improvement activities should be a routine activity for all employees
(Clark, 2016; Liker, & Franz, 2012; Tatham, 2008; Allen, 2010; Kavangh, & Cole, 2013) and a
central aspect of the organization (Liker, & Morgan, 2006; Peterson, 2010; Jusko, 2011; Smith,
2012; Alves, Dinis-Carvalho, & Sousa, 2012; Andel, 2013; Bortolotti et al., 2015). These
findings stay in contrast to Raines (2011), who suggested a formal suggestion system.
Mehri (2006) cautioned that the continuous improvement process stops truly creative new
solutions to problems and employees develop a lot of depth but not the breadth of their
knowledge. The depth of employee knowledge is confirmed by Liker, Morgan (2006).
The engagement of management can be measured by employees if barriers get removed
(Robinson, & Kirsch, 2015; Kavanagh, & Cole, 2013; Netland, 2016) and by the speed of the
implementation process (Vago et al., 2016).
The innovation should not only be driven by standard improvement processes but also
through non-financial goals and key areas set by leadership (Hillberg, 2015). This aspect is a
connection between goal setting, hoshin kanri, and continuous improvement and management
needs to support challenging the status quo by setting the right goals (Sarkar, 2011; Netland,
2016).
Full-Range Leadership Model
The culture of lean was described with its two principles, respect, and continuous
improvement (Liker, & Hoseus, 2010). But no measurement of lean leadership exists. The
leadership theory has developed an approach to measure leadership styles (Avolio, & Bass,
1998).
26
Avolio and Bass (1998) developed the full-range leadership model, including following
leadership styles: transformational leadership, transactional leadership, and passive/avoidant
behavior leadership styles. This research study will describe how the Lean Leadership model
might be connected to these leadership styles.
A central aspect of Lean Leadership seems to be developing a relationship with
employees by being visible through gemba, kaizen, and coaching employees (Flinchbaugh et al.,
2008; Poksinska et al., 2013; Emiliani, & Emiliani, 2013; Thomas, 2016; Delisle, & Turner,
2010). This visibility is related to the transformational leadership style.
The importance of transformational leadership is supported by Poksinska, Swartling, and
Drotz (2013) and Liker, and Franz (2012). In a survey of six companies and 240 respondents, Li,
Nahm, Wyland, Ke, and Yan (2015) concluded, transformational leadership creates employee's
trust into leadership - creating an organizational culture of open communication which leads to
worker participation in problem-solving.
Hartwell and Roth (2010) supported a people-oriented non-competitive leadership style.
In contradiction, Doss and Orr (2007) reported dictatorial/autocratic behaviors in Hungarian
supposed to be lean companies, but their study relied on a self-definition of lean, which is not
congruent with the lean definition of other authors. The people orientation supports the idea of
transformational leadership.
It is important to observe that the lean literature mentions transformational leadership as a
desirable leadership type. The literature does not address other leadership styles suggested by
Avolio and Bass. It seems to be promising to measure the leadership style within an organization
as suggested by Avolio and Bass (1998).
27
Theoretical Background
LMX defines the quality of the leader’s and employee’s relationship (e.g. Graen, &
Scandura, 1987). One party offers something valuable to other. Studies showed that followers
engage in behaviors directly related to their leaders (e.g. Wayne, Shore, & Liden, 1997).
Based on the LMX theory, it might be possible that a leadership style influence directly
the variables respect, proactivity, and innovative work behavior. Having this theoretical
approach, three multivariate regressions seemed promising.
POS defines the quality of the employee and organization relationship by measuring the
employee’s believe how much their organization values their contributions and welfare
(Eisenberger, Huntington, Hutchison, & Sowa, 1986). Depending on how an organization treats
and values an employee, the organization can expect a stronger employee’s devotion to achieve
organizational goals (Settoon, Bennett, & Liden, 1996; Wayne, Shore, & Liden, 1997).
An employee in a lean organization might be related to the organization through
proactivity or innovative work behavior. The quality of these relationships might be influenced
by the perceived received respect from a leader. Taking POS into account, it might be the case
that respect is a mediator in the LMX relationship between transformational leadership and either
proactivity or innovative work behavior. This thought leads to simple mediation models.
But Crant, 2000, cautioned that proactive behavior has emerged as a research stream in
the organizational behavior literature without having a single definition or theory. Research was
conducted through a lens of proactivity and initiative (e.g. Deluga, 1998). A common
denominator might be that employees take an active approach towards organizational goals
(Crant, 2000). A visible organizational goal in lean is the second principle of lean, continuous
improvement. In the definition of proactivity innovating is one of four constructs used.
28
Innovative work behavior is entirely focusing on innovating. Additionally, another construct of
proactivity is problem finding. Innovative work behavior cannot happen without agreement on
having at the first place a problem (Imai, 1986). Therefore, it could be that proactivity is
necessary before innovative work behavior can take place. A double mediation analysis is
necessary to verify this idea. The leadership style addressing the 1st principle of lean, respect for
people, is connected through respect and proactivity to the organizational goal innovative work
behavior.
Overview of the Study
The purpose of this study is to verify if a relationship exists between leadership style and
lean, specifically the two principals, respect for people and continuous improvement, expressed
through innovative work behavior and pro-active employees.
The following questions shall be investigated:
The research questions are
1. Do leaders with transformational leadership style score higher on respect of employees
than other leadership styles at manufacturing organizations in North America?
2. Do leaders with transformational leadership style score higher on innovative work
behavior of employees than other leadership styles at manufacturing organizations in
North America?
3. Do leaders with transformational leadership style score higher on proactivity of
employees than other leadership styles at manufacturing organizations in North America?
4. Does a mediation effect exist between transformational leadership and innovative work
behavior mediated by respect?
29
5. Does a mediation effect exist between transformational leadership and proactivity
mediated by respect?
6. Do double mediations effects exist between one of the five leadership styles and
innovative work behavior mediated by respect and proactivity? (This mean, is the
independent variable, a leadership style, mediated by two variables, respect and
proactivity, influencing the dependent variable, innovative work behavior?)
An improved understanding of these relationships might increase the ability of companies
to introduce and sustain lean management (Bashin, 2012b; Mann 2009).
Based on the questions and the literature review following hypotheses are stated:
1. Leaders with transformational leadership style score higher on respect of employees than
other leadership styles at manufacturing organizations in North America.
2. Leaders with transformational leadership style score higher on proactivity of employees
than other leadership styles at manufacturing organizations in North America.
3. Leaders with transformational leadership style score higher on innovative work behavior
of employees than other leadership styles at manufacturing organizations in North
America.
4. A mediation effect exists between transformational leadership and innovative work
behavior mediated by respect.
5. A mediation effect exists between transformational leadership and proactivity mediated
by respect.
6. Double mediation effects exist between one of the five leadership styles and innovative
work behavior mediated by respect and proactivity.
30
CHAPTER 3: Methodology
Introduction
The current study aims to examine the relationship between leadership styles and respect,
innovative work behavior, and proactivity. An improved understanding of these relationships
might increase the ability of companies to introduce and sustain lean management (Bashin,
2012b; Mann, 2009). The participants are in a leadership role of manufacturing organizations in
North America. This chapter explains the research questions, the research design, the variables,
and the hypotheses,
The research questions are
1. Do leaders with transformational leadership style (TL) score higher on respect of
employees than other leadership styles (LS) at manufacturing organizations in North
America?
2. Do leaders with transformational leadership style score higher on proactivity of
employees than other leadership styles at manufacturing organizations in North America?
3. Do leaders with transformational leadership style score higher on innovative work
behavior (IWB) of employees than other leadership styles at manufacturing organizations
in North America?
4. Does a mediation effect exist between transformational leadership and innovative work
behavior mediated by respect?
5. Does a mediation effect exist between transformational leadership and proactivity
mediated by respect?
6. Do double mediation effects exist between one of the five leadership styles and
innovative work behavior mediated by respect and proactivity?
31
The research questions, corresponding designs, variables, analytical designs, and data
sources are summarized in the following table.
32
Research Question
Study design
Variable
Data
Analysis
Data
Source
Do leaders with
transformational leadership
style score higher on respect of
employees than other leadership
styles at manufacturing
organizations in North
America?
Correlational
Design
LS and
Respect
Multi-variate
regression
Survey
Do leaders with
transformational leadership
style score higher on innovative
work behavior of employees
than other leadership styles at
manufacturing organizations in
North America?
Correlational
Design
LS and
IWB
Multi-variate
regression
Survey
Do leaders with
transformational leadership
style score higher on proactivity
of employees than other
leadership styles at
manufacturing organizations in
North America?
Correlational
Design
LS and
Proactivity
Multi-variate
regression
Survey
Does a mediation effect exist
between transformational
leadership and innovative work
behavior mediated by respect?
Correlational
Design
TL,
Respect,
and
IWB
Multi-variate
regression
meditation
analysis
Survey
Does a mediation effect exist
between transformational
leadership and proactivity
mediated by respect?
Correlational
Design
TL,
Respect,
and
Proactivity
Multi-variate
regression
mediation
analysis
Survey
Do double mediation effects
exist between one of the five
leadership styles and innovative
work behavior mediated by
respect and proactivity?
Correlational
Design
TL,
Respect,
Proactivity,
and
IWB
Multi-variate
regression
double
mediation
analysis
Survey
Table 1: Research Design Summary
Leadership Style (LS), Transformational Leadership (TL), Innovative Work Behavior (IWB)
33
Research Design
The suggested research has six major objectives: to examine the relationship
between (1) leadership styles and respect, and (2) leadership styles and innovative work
behavior, (3) leadership styles and proactivity, to verify (4) a mediated relationship
between transformational leadership to proactivity mediated by respect, (5) a mediated
relationship between transformational leadership to innovative work behavior mediated
by respect, and 6) double mediated relationships between each leadership style to
innovative work behavior mediated by respect and proactivity. A quantitative study,
correlational design (Creswell & Clark, 2010), was used to address these objectives. First,
I present the participants of the study. Second, I explain the measurements used. Third I
explain the data preparation process to synthesize the raw data of the survey to a
meaningful analyzable data set. And finally, I explain the data analysis process. The
analyses process distinguishes between a multivariate regression, mediation analysis, and
double mediation analysis.
Participants.
An eight billion and a four billion company participated in this research study. All
businesses run more than 25 plants each with hundreds of leaders in total. All
organizations are manufacturer converting raw materials into products and face similar
types of problems even though they are active in different markets. Both organizations
provided data about five plants with five to six leaders each. All plants are located in
North America.
The participating organizations were offered to receive a general report about the
current existing leadership styles within their plants, a benchmark to other organizations,
34
and a follow-up discussion for leadership development if desired. The participants taking
the survey were not offered anything.
The participants were employees in leadership positions of the manufacturing
organizations mentioned above. All participants had a mid-level leadership position.
They were selected to inform this research as the leadership team of a plant should drive
the daily continuous improvement efforts. A mid-level leadership position was defined as
having directly reporting employees and not being the top-leader supervising several
manufacturing sites. All participants are leaders and therefore qualified to have their
leadership style measured. Both companies produce goods for industrial clients who are
active in consumer markets. And a competitive environment demands innovation and
proactivity (R. Skillmann, personal communication, January 12th, 2017). Kodak is an
example what happens to even great companies if innovation is ignored. The
organizations conducted for the first time this kind of survey.
Measures.
First, I present the leadership style variables, which is explained by five variables.
A set of five variables are connected to the transformational leadership style (Avolio, &
Bass, 1998), two variables are connected to a transactional leadership style (Avolio, &
Bass, 1998), and two variables are connected to the passive/avoidant behavior style
(Avolio, & Bass, 1998). All five variables together represent the full range leadership
style model (Avolio, & Bass, 1998). Each variable has five items, and the scale per item
is 0 to 4. (0 – not at all, 1 – once in a while, 2 – sometimes, 3 – fairly often, 4 – frequently
if not always).
35
Leadership Style (Independent Variables).
During the literature review the leadership style transformational leadership was
identified as a potential leadership style within lean companies. But transformational
leadership is only one possible style measured by the Multifactor Leadership
Questionnaire (MLQ) (Avolio & Bass, 1998). The MLQ measures the following
leadership styles: Transformational Leadership Style, Transactional Leadership Style, and
Passive/Avoidant Behavior Leadership style (Avolio, & Bass, 1998). All following
variables described are measured on a Likert-type scale, with the following values: 0 (not
at all), 1 (once in a while), 2 (sometimes), 3 (fairly often), and 4 (frequently, if not
always) (Avolio, & Bass, 1998). Following the Multifactor Leadership Questionnaire
(MLQ) (Avolio & Bass, 1998), a complete list of items can only be provided in the IRB
proposal, but sample items will be given below. The MLQ uses five items to measure
each leadership style.
Transformation Leadership.
The Transformational Leadership Style (TL) describes leaders who change
employees’ awareness of essential items and open their mind to evaluate themselves and
the opportunities of their environment (Avolio, & Bass, 1998). Transformational Leaders
seek to improve individuals, groups, and organizations (Avolio, & Bass, 1998). This
leadership style is expressed through five variables (Avolio, & Bass, 1998).
Idealized Influence Idealized Attributes.
Leaders create pride in others for being connected to them, they put the group
before their self-interest, their actions create respect, and they display power
and confidence (Avolio, & Bass, 1998).
36
A sample item is
The person I am rating goes beyond self-interest for the good of the group
Idealized Influence Idealized Behaviors.
Leaders share their most important values and beliefs, they consider morals
and ethics and emphasize the collective mission (Avolio, & Bass, 1998).
A sample item is
The person I am rating considers the moral and ethical consequences of
decisions
Inspirational Motivation.
Leaders are optimistic about the future; they engage employees to accomplish
goals, they communicate a desirable vision and demonstrate confidence that
goals will be met (Avolio, & Bass, 1998).
A sample item is
The person I am rating articulates a compelling vision of the future
Intellectual Stimulation.
Leaders verify assumptions to questions; they verify multiple perspectives
during problem-solving, they make others to look at problems, too and
promote new techniques to accomplish goals (Avolio, & Bass, 1998).
A sample item is
The person I am rating gets me to look at problems from different angles
37
Individual Consideration.
Leaders are coaches; they see employees as individuals, they accept that
individuals have different needs, and aspirations than a group and they help
employees to develop their strengths (Avolio, & Bass, 1998).
A sample item is
The person I am rating helps to develop my strengths
Transactional Leadership Style.
The Transactional Leadership Style describes leaders who use constructive and
corrective transactions (Avolio, & Bass, 1998). The constructive style is called contingent
reward (CR) and the corrective style management-by-exemption (active) (MBEA)
(Avolio, & Bass, 1998).
Contingent Reward.
Leaders who assistant employees in exchange for their efforts, they give clear
performance goals, explain the rewards when goals are achieved, and applaud
when expectations are met (Avolio, & Bass, 1998).
A sample item is
The person I am rating discusses in specific terms who is responsible for
achieving performance targets
Management-by-Exemption (Active).
Leaders who focus on mistakes, and deviations from standards. They
concentrate on dealing with failures and track all mistakes to achieve goals
(Avolio, & Bass, 1998).
38
A sample item is
The person I am rating focuses attention on irregularities, mistakes,
exceptions, and deviations from standards
Passive /Avoidant Behavior Leadership Style.
The Passive/Avoidant Behavior Leadership Style describes leaders who are more
passive and reactive (Avolio, & Bass, 1998). Two variables describe this style,
Management by Exemption (Passive) (MBEP) and Laissez Fair (LF) (Avolio, & Bass,
1998).
Management-by-Exemption (Passive).
Leaders who do not act until a problem is serious. They show a belief in the
status-quo (Avolio, & Bass, 1998).
A sample item is
The person I am rating fails to intervene until problems become serious
Laissez-Faire.
Leaders who avoid getting involved. They make themselves invisible when
needed and avoid making any decisions or delay responding to questions
(Avolio, & Bass, 1998).
A sample item is
The person I am rating avoids getting involved when important issues arise
The MLQ (Form 5X)’s Cronbach’s α varies between .91 and .95 (Avolio & Bass,
2004), which indicates a high level of reliability (Cronbach, 1951).
39
The MLQ (Form 5X) was developed in response to criticism to a formerly used
questionnaire (Bass, & Avolio, 1993). First, Avolio and Bass (1998) conducted a series
of factor analyses of the MLQ 5R, to identify items with the best convergent and
discriminant validities. Second, they (Avolio, & Bass, 1998) used a partial least square
analysis to select items to be included in the MLQ 5X. Third, they (Avolio, & Bass,
1998) developed new items to distinguish between charismatic and transformational
leadership based on a literature review. Fourth, six scholars of leadership received the
MLQ 5X version and made recommendations for item inclusion or exclusion (Avolio, &
Bass, 1998). Further, they (Avolio, & Bass, 1998) conducted a series of Confirmatory
Factor Analyses and LISREL analytics to identify four items per leadership variable to
represent the nine leadership factors (Avolio, & Bass, 1991).
Based on a 2003 normative sample, a confirmatory factor analysis was run
(Antonakis, Avolio, & Sivrasubramaniam, 2003). The validity of the nine-factor
leadership model was supported (Antonakis et al., 2003). The model was also rated stable
within homogenous contexts with following factors: environmental risk, leader-follower
gender, gender of leader, and leader hierarchical level (Antonakis et al., 2003).
The database of the current MLQ 5X was used to further analyze data (Avolio, &
Bass, 1998). A high correlation between transformational and contingent reward
leadership was found (Avolio, & Bass, 1998). The correlations found between
Management by exception active is low positive or negative with transformational and
contingent reward leadership (Avolio, & Bass, 1998). The correlation found between
management by exception passive as laissez-fair and transformational and contingent
reward leadership are negative (Avolio, & Bass, 1998).
40
The validity of transformational leader behavior was supported by Carless (1998).
Rowold and Heinitz (2007) confirmed convergent validity between charismatic and
transformational leadership, but they report as a key finding that charismatic leadership
style does not augment transactional leadership and transformational leadership does.
Also, their study provided evidence that transformational leadership explains
performance ratings over and above transactional leadership. Muenjohn and Armstrong
(2008) conducted a confirmatory factor analysis and stated that the nine-factor leadership
model might be the best to capture the construct factors of transformational and
transactional leadership.
Respect (Dependent Variable Question 1 and Mediating Variable Questions 4, 5,
and 6).
Van Quaquebeke and Eckloff (2010) defined respect as an attitude to others
justifying attention and behavior which returns in the targeted person appreciation in
worth and importance. During the literature review, the lean principle of respect for
people was documented. As this principle seems to be a foundational element of lean, it
was measured how subordinates feel respected by their superiors.
Based on their definition, Van Quaquebeke and Eckloff (2010) developed a 12-
item survey by identifying 149 statements and creating 19 categories through conducting
qualitative research based on 426 participants. Additionally, they conducted two more
studies to empirically derive feasible measurement items (Van Quaquebeke, & Eckloff,
2010). They investigated all items of the 149 statements with a minimum of ranking 6
(scale 1 to 7) and with a high correlation to the respectful leadership marker (Van
Quaquebeke, & Eckloff, 2010). After finding 12 items, they investigated the
41
psychometric qualities of them (Van Quaquebeke, & Eckloff, 2010). They conducted a
principal component factor analysis, discriminatory power cores were calculated, and
they claimed that their survey assesses the construct of respectful leadership well (Van
Quaquebeke, & Eckloff, 2010). They reported a Cronbach’s α between .95 and .96,
which indicates a high level of reliability (Cronbach, 1951).
The Respectful Leadership Scale developed by van Quaquebeke and Eckloff
(2010) is rated on a 5-point Likert scale from 0 (not at all), 1 (once in a while), 2
(sometimes), 3 (fairly often), and 4 (frequently, if not always). A sample item is (all items
Appendix A);
Trust my ability to independently and self-reliantly perform well
Wong, Tjosvold, and Khong (2016) suggested that respectful leaders enable
followers to manage conflicts cooperatively, which supports construct validity.
A limitation in the validity of this construct is that Van Quaquebeke’s and
Eckloff’s (2010) studies were conducted in Germany. Other cultures might define
respectful leadership differently. In this study organizations in North America shall be
measured. These are western cultures and following Hofstede (2001) they seem to be
comparable. Power distance was 35 for Germany and 40 for the USA, Masculinity was
66 for Germany and 62 for the USA, Uncertainty Avoidance was 65 for Germany and 46
for the USA, and Individualism was 67 for Germany and 91 for the USA.
Proactivity (Dependent Variable Questions 3, 5 and Mediating Variable Question 6).
Bandura (2001) suggested that human activity is intentionally driven by action
and individuals have the opportunity to change their surroundings. Based on Bandura’s
42
thought, Bateman and Crant (1993) defined a proactive personality scale as the ability to
intentionally and directly change current circumstances, social or nonsocial. Proactive
individuals might exhibit behaviors of problem finding and idea championing (Maynes,
McCall & Kaplan, 1985), innovating (Hirschman, 1970), task revision (Staw, & Boettger,
1990), and affect change (Grant, & Ashford, 2008). The second basic principle of lean is
continuous improvement. An assumption was that proactive individuals are needed to
achieve continuous improvement as the later requires constant change and the former
seems to foster the same. Leadership style might allow proactive employees or not.
Bateman and Crant (1993) developed a 17-item proactive scale and report a
Cronbach’s α = .89, which indicates a good level of reliability (Cronbach, 1951).
Seibert, Crant, and Kraimer (1999) used a modified proactive personality scale
and reported a Cronbach’s alpha of .85. They verified the validity of the shortened 10-
item scale through a pre-test of 181 MBA undergrad students. After eliminating seven
items from the scale, they reported a difference of Cronbach’s alpha of 0.02 and
concluded that the shortened version appeared to be comparable. The shortened scale has
been used in this study to ease the burden of the study subjects to answer an already long
survey.
The proactive scale uses a Likert scale from 1 to 5. Because other data requires a
different Likert scale the following scale shall be used: 0 to 4, step 1 from (0) strongly
disagree, (1) disagree, (2) neutral, (3) agree, (4) strongly agree.
A sample item is (all items Appendix A)
When I have a problem, I tackle it head-on
43
Bateman and Crant (1993) conducted three studies to assess the scale’s
psychometric properties. Factor analysis supported the uni-dimensionality of the scale
(Bateman, & Crant, 1993). A moderate correlation with the need for achievement and
dominance supported a convergent validity (Bateman, & Crant, 1993). Locus of control
was not significantly correlated to proactive personality supporting discriminant validity
(Bateman, & Crant, 1993). A limiting factor might be that the participants in their
samples were all students even though in one sample the participants were MBA
students. In the suggested study the participants will be employees who may or may not
have studied and likely are from a different age range. They reported discriminant
validity between the proactive scale and age, sex, years of work experience, locus of
control, private self-consciousness, intelligence, agreeableness, neuroticism, and
openness (Bateman, & Crant, 1993).
The proactive personality scale was used in a research model to connect
proactivity personality and job search within graduating college students, but the validity
of the proactive personality scale was not discussed. (Brown, Kane, Cober, Levy, &
Shalhoop, 2006).
Crant (1995) provided evidence for the criterion validity of the proactive
personality scale by conducting a hierarchical regression with several control variables.
The proactive personality scale provided a significant 8% variance explanation (Crant,
1995).
Eby, Butts, and Lockwood, 2003, conducted a study using following predictors:
proactive personality, openness to experience, career insight, experience with a mentor,
internal networks, external networks, career/job-related skills, and career identity. All
44
predictors were significantly correlated to three criteria, further supporting criterion
validity (Eby, Butts, & Lockwood, 2003).
Seibert, Kraimer, and Crant, 2001, used the shortened 10-item scale in a
longitudinal study to research the connection between proactive personality and salary
progression, promotions in past two years, and career satisfaction in a structured equation
model analysis. They reported significant relations, further supporting the criterion
validity of the proactive personality scale (Seibert, Kraimer, & Crant, 2001).
Zhou and Shi (2009) discussed the construct validity by focusing on validating a
Chinese translation of the 10-item scale. They used following measurements: proactive
personality scale, big five factors, political skill, career satisfaction, job performance, and
general self-efficacy (Zhou, & Shi, 2009). They conducted an exploratory factor and
confirmatory factors analysis and concluded that the one-factor model is acceptable for
the 10-item scale (Zhou, & Shi, 2009). They computed Pearson correlation coefficients
between the total scores of the Chinese translation and other related measures and
conducted two-tailed significant tests to confirm construct validity (Zhou, & Shi, 2009).
They conducted a hierarchical regression to verify the ability of proactive personality to
predict career satisfaction and job performance and report support for criterion validity
(Zhou, & Shi, 2009).
Yang and Chau (2016) conducted a study about the relationship between
proactive personality and career success mediated by the leader-member exchange. They
used hierarchical regression, a moderated path analysis, and a Monte-Carlo simulation to
confirm the relationship (Yang, & Chau, 2016) providing further evidence of criterion
validity.
45
Innovative Work Behavior (Dependent Variable Questions 2, 4, and 6)
Innovative Work Behavior (IWB) is defined through individual employees who
“develop, carry, react to, and modify ideas” (Van de Ven, 1986, p. 592) and the root
source, ideas (Scott, & Bruce, 1994). Janssen (2000) developed a 9-item scale based on
the works of Scott and Bruce (1994), individual innovative behavior, and Moss Kanter
(1988), stages of innovation. Janssen (2000) suggested three items per each concept, idea
promotion, idea generation, and idea realization. Janssen (2000) distinguished between
“self-rated” and “leader-rated” scores and preferred to only use the “self-rated” scores
due to following three reasons: 1. The cognitive representation and reports on ones on
Innovative Work Behavior might be better due to the knowledge about one’s own work.
2. The assessment of the IWB is a subjective performance appraisal. 3. Supervisors might
miss genuine innovative activities not reported. Intercorrelation between the three
concepts of IWB was reported to be between .76 and .85 for self-reports (Janssen, 2000).
During the literature review, constructs were identified, which foster innovative
work behavior. These are 1. Asking why not who, trying to promote innovation through
identifying root problems and therefore generating ideas to solve the root problem. 2.
Training of employees about problem-solving tools, employees shall be enabled to
innovate in a constructive manner learning new ways to analyze and solve problems.
Janssen’s (2000) 9-item scale has nine items on a Likert scale from 1 = never to 7
= always. Because other data requires a different Likert scale the following scale shall be
used: 0 to 4 step 1 from (0) never, (1) rarely, (2) sometimes, (3) frequently, to (4) always.
An example item is (all items Index A);
Do you create new ideas for difficult issues?
46
Ramamoorty, Flood, Slattery, and Sardessai (2005) used the Janssen scale but
modified the scale to a five-point Likert scale. They developed a model connecting
meritocracy, justice perceptions, pay, job autonomy with met expectations, obligation to
innovate with IWB and used Padhazur’s method to derive path coefficients and reported
significant relationships between constructs (Ramamoorty, Flood, Slattery, & Sardessai,
2005). Their research supports criterion validity.
Stock (2015) used a reduced scale of 5-items due to the service environment of
her study and excluded items related to idea promotion. To verify the reliability and
validity, an exploratory and confirmatory factor analysis were conducted; Cronbach's α
exceeding .7, passing recommended values (Bagozzi, Yi, & Philipps, 1991; Nunnally, &
Bernstein, 1978) and a composite reliability of greater than .6, passing the threshold value
(Bagozzi, Yi, & Philipps, 1991), were reported (Stock, 2015). Stock (2015) reported a
pass for a test of discriminant validity following Fornell and Larcker’s (1981) criterion.
Janssen’s (2000) reported a Cronbach's α = .95, which indicates a high level of
reliability (Cronbach, 1951). For another scale based on Scott and Bruce (1994) work, a
Cronbach's α = .89 (Scott, & Bruce, 1994), a Cronbach's α =.76 (Chen, & Aryee, 2007),
and a Cronbach's α =.87 (supervisor rating) / .86 (self-rating) (Carmeli, Meitar, &
Weisberg, 2006) were reported. These results indicate that Janssen’s scale seems to be
better.
Organization and Gender (Control Variable).
A nominal variable Organization was used to account for possible within-
organization effects on the independent variables. Schein (1985b) proposed that an
organization has an influence on the culture of a unit, and Mann (2008) proposed that the
47
top leader of an organization influences the culture of the same. As this study will focus
on the unit level, the organization’s influence shall be controlled.
A nominal variable gender (female =1 / male = 0) will be used to account for
possible gender effects on the independent variables. Avolio and Bass (1998) stated that
gender might affect the leadership style variable.
Reliability of Scales.
For the three scales not so frequently used, Respect, Proactivity, and Innovative
Work Behavior, I used Cronbach's alpha method (Cronbach, 1951) to estimate the
reliability of the scales used. For values above 0.9, the reliability is excellent, between 0.8
and 0.9 good, and between 0.7 and 0.8 acceptable.
Procedure
Preparation of Survey.
The company Mind Garden agreed to enhance the MLQ-survey by the three
suggested surveys against payment. Therefore, only one survey with all items was
deployed instead of four different surveys.
Data Collection.
The participating organizations had defined the to-be-analyzed leaders and their
hierarchical level. The corresponding empty table to collect data is attached in Appendix
B.
The survey was administered through the MLQ 5X software TransformTM. A
customization of the MLQ 5X was done to include the three other instruments to the
48
MLQ 5X survey tool (Avolio, & Bass, 1998; Respectful Leadership Sale tool (Van
Quaquebeke, & Eckloff, 2010; see Appendix A); Innovative Work Behavior tool
(Janssen, 2000; see Appendix A); Proactive scale tool (Bateman, & Crant, 1993; see
Appendix A). Additionally, the survey was translated by language experts into German
and Spanish.
The identified leaders received a leadership MLQ. The leaders sent via the MLQ
online tool emails to their superior(s), same level colleagues, and their subordinates. Then
all these contacted employees answered the enhanced MLQ regarding the leadership style
of the leader. This technique is called 360-degree feedback.
MLQ 5X software TransformTM provided the data values for all variables per
survey participant. The software aggregated the data per leader for all leadership style
variables. The data for Respect, Proactivity and IWB were collected for all leaders by the
subordinates, pears, and superiors identified by them.
Data Preparation.
The participating leaders were separated into three levels: top-level business unit
leaders, low-level leaders with direct supervision of employees and workers but no
indirect supervised personnel, and mid-level, all other leaders. Only mid-level leaders
were included in this study.
Only if a minimum of three other people rated the leader on the MLQ 5X, the
leader’s data was used in the data analysis section (Avolio, & Bass, 1998). Who these
three people were was irrelevant as long as they were directly connected to the leader
(Avolio, & Bass, 1998). The reason for setting the threshold at three is that Avolio and
Bass (1998) present such a ratio for the construct validity of the MLQ 5X.
49
Only if a minimum of three subordinates participated in the enhanced MLQ 5X of
the leader, the variables Respect, Innovative Work Behavior, and Proactivity were
connected to the leader’s leadership style. For the measurement of the dependent
variables, no minimum amount of answers is suggested by the respective authors, but the
same criterion as for the MLQ seemed reasonable. The middle leaders received five
leadership style scores through the MLQ. The scores for Respec, Proactivity, and
Innovative Work Behavior were calculated by taking the average of only the
subordinates.
Data Outlier Analysis.
A test for outliers was conducted by checking three criteria (Cohen, Cohen, West
& Aiken, 2013): 1. Leverage, 2. Influence on Data and 3. Discrepancy.
1. Leverage: SPSS reported the centered leverage value (Cohen et al., 2013).
Due to the number of potential participants of over 30, I assumed a small size
sample. This allows a cut-off value of 3k/n with k number of independent
variables and n number of cases (Cohen et al., 2013). The identified cases
might be outliers.
2. Influence of data: I will use Cook’s Di to identify the global influence of a
case (Cohen et al., 2013). Assuming a small sample again, the cut-off value
for Cook’s distance is> 1 (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
3. Discrepancy: Externally studentized residuals is a preferred measure of
discrepancy (Cohen et al., 2013). SPSS terminology for externally studentized
residuals is “studentized Deleted Residual” (Cohen et al., 2013). Following
Cohen, Cohen, West, and Aiken (2013) data analysts used cut-off values from
50
2.0 to 4.0 in large samples to reduce the number of potential outliers. I will
keep the number of possible outliers below 3.
I plotted Cook’s D against the centered leverage value while identifying each data
point with a label. If Data points are singular Cook’s D scale and/or above the calculated
threshold for centered leverage value, I considered the data point as a potential outlier
(Cohen et al., 2013). I plotted “Case ID” and “Studentized Deleted Residual” with SPSS.
By looking at the cut-off line, I identified potential outliers (Cohen et al., 2013).
If a data point passes all three criteria, I excluded it from the statistical analysis. If
one data point passed two tests in a rather strong manner, I paid special attention to it and
decided case-by-case if it was excluded or not.
Regression Assumption Analysis.
Now I could start to verify the assumptions for regression:
1. Normality of the dependent variable
I used the Kolmogorov-Smirnov and Shapiro-Wilk tests to verify this assumption
(Lomax & Hahs-Vaughn, 2012).
2. Independence
I verified the residual plots of studentized residuals with unstandardized predicted
values and all used independent variables. No pattern should be visible, and the
values should fall within a band from -2 to 2 (Lomax & Hahs-Vaughn, 2012).
There is a risk of violating this assumption in the suggested study as data was
gathered in blocks (plants). As a criterion, I used the Durbin-Watson test. The
results should not be below 1 or above 3.
51
3. Homogeneity of Variance
The created plots were used to verify the spread of the studentized residuals,
which should be fairly constant (Lomax & Hahs-Vaughn, 2012).
4. Linearity
I produced multiple scatter plots between all variables. By looking at the
produced scatter plots, I was able to confirm this assumption if I observed only
linear or random patterns (Lomax & Hahs-Vaughn, 2012). Additionally, I verified
the residual plots again by observing the lowess line, which should show no
pattern (Cohen et al., 2013).
5. Normality
I plotted the Q-Q scatterplot of unstandardized residuals. The points should fall on
along a straight diagonal line (Lomax & Hahs-Vaughn, 2012). I noted the
descriptive statistics. The skewness and kurtosis of the data should be outside the
range from -2 to 2 (Lomax & Hahs-Vaughn, 2012). I printed the frequency
histogram, and a normal distribution should be observable (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013). I created a box-plot and used the tests of
normality from Kolmogorov-Smirnov and Shapiro-Wilk to verify the normality
assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
6. Fixed X
As I excluded missing data, all x values had fixed values, and all predicted data
will be only on the scale from 0 to 4. Thus, the assumption was passed (Lomax &
Hahs-Vaughn, 2012).
7. Multicollinearity
52
The VIF should be lower than 10 (Lomax & Hahs-Vaughn, 2012). However,
Cohen at al. (2013) suggested that this general belief might be too high for
behavioral science, yet they do not provide any other threshold. Additionally, I
reviewed the corresponding regression model(s) if the eigenvalues of the
predictors were close to 0. If the model consisting of the predictor variable with
low eigenvalue as the dependent variable and the other predictor variable(s) as the
independent variable(s) produced an R2 below 0.9, it suggests that
multicollinearity is not an issue (Lomax & Hahs-Vaughn, 2012).
Data Analysis
Quantitative analyses were conducted to describe the relationships between the
depended and independent variables. First multiple regressions based on the five
leadership style variables and the three depended variables were conducted. Then two
mediation models and five double mediation analysis were analyzed.
Multiple Regression Models.
The first three research questions required multi-variate regressions.
1. Do leaders with transformational leadership style score higher on respect than
other leadership styles at manufacturing organizations in North America?
H10: There is no statistically significant relationship between the leadership style
and respect.
H1a: There is a statistically significant relationship between the leadership style
and respect.
The basic model is
53
Leadership Style Respect
The multiple linear regression model is (Lomax & Hahs-Vaughn, 2012)
𝑅𝑒𝑠𝑝𝑒𝑐𝑡 = 𝛽𝑅𝑒𝑠𝑝𝑒𝑐𝑡𝑗𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗𝑖+ 𝛼𝑅𝑒𝑠𝑝𝑒𝑐𝑡 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 + 𝜀𝑖
Where
𝑅𝑒𝑠𝑝𝑒𝑐𝑡 is the criterion variable
𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗 are predictor variables (TL, CR, MBEA, MBEP, LF)
𝛽𝑅𝑒𝑠𝑝𝑒𝑐𝑡𝑗 are the slopes for 𝑅𝑒𝑠𝑝𝑒𝑐𝑡 predicted by 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 𝑗
𝛼𝑅𝑒𝑠𝑝𝑒𝑐𝑡 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 is the intercept for 𝑅𝑒𝑠𝑝𝑒𝑐𝑡 predicted by
𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒
𝜀𝑖 are the population residuals or errors of prediction
i represents an index for a particular case with 1 to n participants in this study
j represents an index for the five different leadership styles in this study.
According to Pedhazur (1997), it is preferable to use the non-standardized values
for the variables to receive a more consistent prediction. The least square criterion is used
to find the regression line. With the help of an F-Statistic, the significance of the model
was verified. The multiple r squared value was verified to indicate how much variation
was predicted by the x variables. The significance of the slopes’ coefficients were
verified by using a t-statistic. The unstandardized and standardized slopes, the confidence
interval (CI) around the unstandardized slope, the intercept and its significance (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013), and the estimated power as calculated by
54
G*Power 3.1.9.2 (Faul, Erdfelder, Buchner, & Lang, 2007) were noted. With the help of
this verification, the hypothesis was rejected or not rejected.
2. Do leaders with transformational leadership styles score higher on innovative
work behavior (IWB) than other leadership styles at manufacturing
organizations in North America?
H20: There is no statistically significant relationship between the leadership styles
and innovative work behavior.
H2a: There is a statistically significant relationship between the leadership styles
and innovative work behavior.
The model is
Leadership Style Innovative Work Behavior (IWB)
The multiple linear regression model is (Lomax & Hahs-Vaughn, 2012)
𝐼𝑊𝐵 = 𝛽𝐼𝑊𝐵𝑗𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗𝑖+ 𝛼𝐼𝑊𝐵 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 + 𝜀𝑖
Where
𝐼𝑊𝐵 is the criterion variable
𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗 are predictor variables
𝛽𝐼𝑊𝐵𝑗 are the slopes for 𝐼𝑊𝐵 predicted by 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗
𝛼𝐼𝑊𝐵 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 is the intercept for 𝐼𝑊𝐵 predicted by 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒
𝜀𝑖 are the population residuals or errors of prediction
i represents an index for a particular case with 1 to n participants in this study
55
j represents an index for the five different leadership styles in this study.
According to Pedhazur (1997), it is preferable to use the non-standardized values
for the variables to receive a more consistent prediction. The least square criterion is used
to find the regression line. With the help of an F-Statistic, the significance of the model
was verified. The multiple r squared value was verified to indicate how much variation
was predicted by the x variables. The significance of the slopes’ coefficients were
verified by using a t-statistic. The unstandardized and standardized slopes, the confidence
interval (CI) around the unstandardized slope, the intercept and its significance (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013), and the estimated power as calculated by
G*Power 3.1.9.2 (Faul et al., 2007) were noted. With the help of this verification, the
hypothesis was rejected or not rejected.
3. Do leaders with transformational leadership style score higher on proactivity of
employees than other leadership styles at manufacturing organizations in North
America?
H30: There is no statistically significant relationship between the leadership style
and proactivity of employees.
H3a: There is a statistically significant relationship between the leadership style
and proactivity of employees.
The model is
Leadership Style Proactivity
The multiple linear regression model is (Lomax & Hahs-Vaughn, 2012)
𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 = 𝛽𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦𝑗𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗𝑖+ 𝛼Proactivity 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 + 𝜀𝑖
56
Where
𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 is the criterion variable
𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗 are predictor variables
𝛽𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦𝑗 are the slopes for 𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 predicted by 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗
𝛼𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒 is the intercept for 𝑃𝑟𝑜𝑎𝑐𝑡𝑖𝑣𝑖𝑡𝑦 predicted by
𝐿𝑒𝑎𝑑𝑒𝑟𝑠ℎ𝑖𝑝 𝑆𝑡𝑦𝑙𝑒𝑗
𝜀𝑖 are the population residuals or errors of prediction
i represents an index for a particular case with 1 to n participants in this study
j represents an index for the five different leadership styles in this study.
According to Pedhazur (1997), it is preferable to use the non-standardized values
for the variables to receive a more consistent prediction. The least square criterion is used
to find the regression line. With the help of an F-Statistic, the significance of the model
was verified. The multiple r squared value was verified to indicate how much variation
was predicted by the x variables. The significance of the slopes’ coefficients were
verified by using a t-statistic. The unstandardized and standardized slopes, the confidence
interval (CI) around the unstandardized slope, the intercept and its significance (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013), and the estimated power as calculated by
G*Power 3.1.9.2 (Faul et al., 2007) were noted. With the help of this verification, the
hypothesis was rejected or not rejected.
57
Mediation Models.
Simple Mediation Models.
For the next questions, the variable Respect was converted into a mediator
variable.
4. Does a mediation effect exist between transformational leadership and proactivity
mediated by respect?
H40: Respect does not mediate the relationship between Transformational
Leadership and Proactivity.
H4a: Respect does mediate the relationship between Transformational Leadership
(TL) and Proactivity.
The model is:
TL Respect Proactivity
The analytical approach is based on the work of Baron and Kenny (1986), who
distinguished between moderators and mediators. They provided the analytic procedure
used here. The mediator respect could be the carrier of the effect leadership style on
proactivity. To do so, I used the SPSS Add-On “Process 3.0” from Andrew F. Hayes and
followed Sobel’s approach (1982) and a bootstrapping technique suggested by Hayes
(2013) to verify the mediation results.
Sobel (1982) suggested a method to assess the significance of indirect effects (in
this case mediation) based on the variance-covariance matrix of the coefficients to
compute the matrix of partial derivatives and the estimated asymptotic variance-
covariance matrix of the indirect effect. Hayes (2017) suggested using random sampling
58
with replacement (aka bootstrapping) to assign a measure of accuracy (confidence
interval) for the indirect effect. Both authors suggested methods to provide evidence if an
indirect effect (mediation) is statistically supportable.
A regression was conducted between TL and Proactivity, a regression between TL
and Respect, and a multiple regression between TL and Respect on Proactivity.
According to Pedhazur (1997), it is preferable to use the non-standardized values for the
variables to receive a more consistent prediction. To find the regression lines, the least
square criterion was used. With the help of an F-Statistic, the model’s significance was
verified (Hayes, 2017). Multiple r squared was noted to indicate how much variation was
predicted by the x variables (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013). The
significance of the slopes coefficients was verified by using a t-statistic (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
The regression between transformational leadership style and proactivity had to
be significant (Hayes, 2017). The regression between Transformational Leadership and
Respect had to be significant (Hayes, 2017). The multiple regression between
transformational leadership style and respect on proactivity had to be significant (Hayes,
2017). The effect of transformational leadership style on proactivity had to be not
significant in the multiple regression (Hayes, 2017). In all cases, the F-test was used to
verify the significance of the regression models and the t-test for the coefficients
significance in the multiple regression (Lomax & Hahs-Vaughn, 2012; Cohen et al.,
2013). Additionally, this procedure allowed to receive the slopes of the regression lines
between transformational leadership and respect, respect and proactivity,
59
transformational leadership and proactivity, and finally the multiple regression between
transformational leadership style and respect on proactivity (Hayes, 2017).
With the help of the slopes and standard errors, a z-score =𝑎𝑏
𝑠𝑎𝑏 was calculated to
test if the indirect effect was significant or not (Sobel, 1982). Additionally, a
bootstrapping approach was used to verify the significance of the mediation model as
suggested by Hayes (2017). If the confidence interval did not include 0, a significant
indirect coefficient was assumed (Hayes, 2017).
5. Does a mediation effect exist between transformational leadership (TL) and
innovative work behavior (IWB) mediated by respect?
H50: Respect does not mediate the relationship between Transformational
Leadership and Innovative Work Behavior.
H5a: Respect does mediate the relationship between Transformational Leadership
and Innovative Work Behavior.
The model is:
TL Respect Innovative Work Behavior
The analytical approach is based on the work of Baron and Kenny, (1986). The
mediator respect could be the carrier of the effect leadership style on Innovative Work
Behavior. To do so, I used the SPSS Add-On “Process 3.0” from Andrew F. Hayes to
follow Sobel’s approach (1982) and a bootstrapping technique suggested by Hayes
(2017).
60
A regression was conducted between TL and Innovative Work Behavior, and
between TL and Respect, and a multiple regression between TL and Respect on
Innovative Work Behavior. According to Pedhazur (1997), it is preferable to use the
non-standardized values for the variables to receive a more consistent prediction.
To find the regression lines, the least square criterion was used. With the help of
an F-Statistic, the model’s significance was verified (Hayes, 2017). Multiple r squared
was noted to indicate how much variation was predicted by the x variables (Lomax &
Hahs-Vaughn, 2012; Cohen et al., 2013). The significance of the slopes coefficients was
verified by using a t-statistic (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
The regression between transformational leadership style and Innovative Work
Behavior had to be significant (Hayes, 2017). The regression between Transformational
Leadership and Respect had to be significant (Hayes, 2017). The multiple regression
between transformational leadership style and respect on Innovative Work Behavior had
to be significant (Hayes, 2017). The effect of transformational leadership style on
Innovative Work Behavior had to be not significant in the multiple regression (Hayes,
2017). In all cases, the F-test was used to verify the significance of the regression models
and the t-test for the coefficients’ significance in the multiple regression (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013). Additionally, this procedure allowed to receive the
slopes of the regression lines between transformational leadership and respect, respect
and Innovative Work Behavior, transformational leadership and Innovative Work
Behavior, and finally the multiple regression between transformational leadership style
and respect on Innovative Work Behavior (Hayes, 2017).
61
With the help of the slopes and standard errors, a z-score =𝑎𝑏
𝑠𝑎𝑏 was calculated to
to test if the indirect effect was significant or not (Sobel, 1982). Additionally, a
bootstrapping approach was conducted to verify the significance of the mediation model
as suggested by Hayes (2017). If the confidence interval did not include 0, a significant
indirect coefficient was assumed (Hayes, 2017).
Double Mediation Models.
Five double mediation models were run based on the Leadership Style (LS)
independent variables. Here only one basic hypothesis is presented using Leadership
Style to represent all five leadership styles presented, Transformational Leadership,
Contingent Reward, Management by Exception Active, Management by Exception
Passive, and Laissez Fair.
6. Do double mediation effects exist between one of the five leadership styles and
innovative work behavior mediated by respect and proactivity?
H60: Respect and Proactivity do not mediate the relationship between Leadership
Style (LS) and Innovative Work Behavior.
H6a: Respect and Proactivity do mediate the relationship between Leadership
Style (LS) and Innovative Work Behavior
The model is:
LS Respect Proactivity Innovative Work Behavior
The analytical approach is based on the work of Baron and Kenny, (1986). The
mediators Respect and Proactivity could be the carriers of the effect Leadership Style on
62
Innovative Work Behavior. To do so, the SPSS Add-On “Process 3.0” from Andrew F.
Hayes was used to follow a bootstrapping technique suggested by Hayes (2017).
A regression was conducted between the independent variable Leadership Style
and the dependent variable Innovative Work Behavior. This regression had to be
significant (Hayes, 2017) to support double mediation. A regression was conducted
between the independent variable Leadership Style and dependent variable Respect. This
regression has to be significant (Hayes, 2017) to support double mediation. According to
Pedhazur (1997), it is preferable to use the non-standardized values for the variables to
receive a more consistent prediction. To find the regression lines, the least square
criterion was used. With the help of an F-Statistic, the significance of the model was
verified (Hayes, 2017). Multiple r squared was noted to indicate how much variation was
predicted by the x variables (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013). The
significance of the slopes coefficients was verified by using a t-statistic (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
A multivariate regression was conducted between the independent variables
Leadership Style, Respect and dependent variable Proactivity. The path between
Leadership Style and Proactivity had to be not significant, but the path between Respect
and Proactivity had to be significant to support double mediation (Hayes, 2017). The F-
test was used to verify the significance of the regression models and the t-test for the
coefficients significance in the multiple regression (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
A multivariate regression was conducted between the independent variables
Leadership Style, Respect, Proactivity and the dependent variable Innovative Work
63
Behavior. The path between Leadership Style and Innovative Work Behavior had to be
not significant to support double mediation (Hayes, 2017). The path between Respect and
Innovative Work Behavior had to be not significant to support double mediation (Hayes,
2017). The path between Proactivity and Innovative Work Behavior has to be significant
to support double mediation (Hayes, 2017). The F-test was used to verify the significance
of the regression models and the t-test for the coefficients’ significance in the multiple
regression (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
A bootstrapping approach was conducted as suggested by Hayes (2017) to verify
the significance of the paths. If the confidence interval of 95% did not include 0, a
significant indirect coefficient was assumed (Hayes, 2017).
Sample Size.
Multiple Regression.
G*Power 3.1.9.2 (Faul et al., 2007) was used to conduct an a priori power
analysis for a linear multiple regression with fixed model and R2 deviation from zero with
5 independent variables. G*Power is a valid method of sample size estimation (Faul et
al., 2007). The power analysis indicated 27 participants would yield an 95% chance
(power = 0.95) of detecting a large effect size of 0.5 (Lomax & Hahs-Vaughn, 2012). A
large effect size was chosen for the study because the theory hints to a strong relationship
between Leadership Style and Respect, Proactivity, Innovative Work Behavior. Also,
multiple studies using the MLQ and the corresponding leadership styles have revealed
large effect sizes regarding other constructs (Avolio, & Bass, 1998). A pilot study was
deemed impractical.
64
Multiple Regression Mediation Models.
G*Power 3.1.9.2 (Faul et al., 2007) was used to conduct an a priori power
analysis for a linear multiple regression with fixed model and R2 deviation from zero with
2 independent variables. G*Power is a valid method of sample size estimation (Faul et
al., 2007). The power analysis indicated 32 participants would yield an 95% chance
(power = 0.95) of detecting a large effect size of 0.35 (Lomax & Hahs-Vaughn, 2012). A
large effect size was chosen for the study because the theory hints to a strong relationship
between Transformational Leadership and Respect on Proactivity as Innovative Work
Behavior for people. Also, multiple studies using the MLQ and the corresponding
leadership styles have revealed large effect sizes regarding other constructs (Avolio, &
Bass, 1998). A pilot study was deemed impractical.
Multiple Regression Double Mediation Models.
G*Power 3.1.9.2 (Faul et al., 2007) was used to conduct an a priori power
analysis for a linear multiple regression with fixed model and R2 deviation from zero with
3 independent variables. G*Power is a valid method of sample size estimation (Faul et
al., 2009). The power analysis indicated 56 participants would yield an 95% chance
(power = 0.95) of detecting a medium effect size of 0.25 (Lomax & Hahs-Vaughn, 2012).
A medium effect size was chosen for the study because the theory hints to a relationship
between Leadership Style on Respect on Proactivity to Innovative Work Behavior. A
pilot study was deemed impractical.
65
Assumptions and Limitations
The researcher assumed that all raters will provide honest responses. This is a
valid assumption as the survey was administered by a third party to all raters so that
anonymity was provided and no retribution needed to be feared.
A limitation was that the participants are not selected randomly.
Ethics of this Study
In research, ethical guidelines principles are respect for people, beneficence,
justice, and non-malfeasance (Fuji, 2012; Muthuswamy, 2013; Wester, 2011). An
informed consent section was included in the survey instrument as the first section, see
appendix C. Only if the participants clicked on the given consent button, the survey had
been administered. Before sending the survey out, the UNM IRB approval was obtained,
see appendix C.
66
CHAPTER 4: ANALYSIS AND RESULTS
Data Collection
A leader was defined as a person who supervises employees and is not
responsible for a multiple plant business unit. The data collected about a leader needed to
be provided by a minimum of three followers, because only then a leadership style can be
defined following the MLQ Handbook.
Organization A identified 30 leaders and Organization B eight, who should have
participated in this research. 30 leaders of Organization A participated, but for four
leaders the minimum threshold of three raters was not achieved. Eight leaders of
Organization B participated, but for two leaders the minimum threshold of three raters
was not achieved. All surveys sent out to the 38 leaders were answered, which is a
success rate of 100%. All leaders were asked to invite their followers, peers, and
supervisors to rate them. They sent out in total 206 surveys. This is called a 360-degree
feedback. However, for this analysis, only the data from the followers were used to rate a
leader. And a leader only was included if a minimum of 3 followers rated the supervisor.
From total 38 potential data points, 32 were achieved due to the minimum threshold of
followers. This is a success rate of 84.21%. For being able to create these 32 data points,
in total 139 surveys were conducted. Given that a total of 206 surveys were conducted,
67.48% were used in the data. As only two out of 32 data points were female, the gender
control variable got dropped.
67
Data Outlier and Assumption Analysis
Three outlier analyses are provided for the multiple regression models, then two
for the mediation models and then five for the double mediation models. Three
verifications were conducted for outliers:
A scatter plot showed Cook’s D (1st verification) and Centered Leverage Values
(2nd verification). Cases which seem to have a larger gap to other cases might be outliers
(Cohen et al., 2013). The Centered Leverage cut off value is 0.46875 (cut off value is
calculated as suggested by Cohen et al., 2013). An analysis of discrepancy (3rd
verification) used a scatter plot between externally studentized residuals (in SPSS
Studentized Deleted Residuals) and Case ID. Attention was paid to cases above 2 or
below -2 as suggested by Cohen et al., 2013.
If the same case was identified in all three verifications, the case was excluded
from the regression analysis. If only two verifications provided positive results, a closer
look at the evidence was taken to decide to keep the data point or not, basically verifying
how strong the violations of these two verifications were.
For the assumptions of regression, six verifications were conducted. 1. If the
dependent variable is normally distributed. 2. If the independence assumption is fulfilled.
3. If the homogeneity of variance assumption is fulfilled. 4. If the linearity assumption is
fulfilled. 5. If the normality assumption is fulfilled. 6. If the multicollinearity assumption
is fulfilled.
Leadership Styles and Respect.
68
Leadership Styles and Respect - Outlier Analysis.
The centered leverage value cut-off value was 0.46875 calculated as described in
Chapter 3. Case 8 might be an outlier as it is on both criteria (Figure 6) noticeable.
.
Figure 6: Leadership Styles and Respect, Outlier Cook’s D / Centered Leverage
Cases 8 and 18 might be outliers as both values are above the 2.0 threshold
(Figure 7).
69
Figure 7: Leadership Styles and Respect, Outlier Externally Studentized residuals
Case 8 was identified three times as a potential outlier. This is the reason why
case 8 was excluded in the following regression analysis.
Leadership Styles and Respect - Assumption Analysis.
Normality of Respect.
Both normality tests (Table 2) were not significant, and the assumption was
reasonably fulfilled (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Respect
.087
31
.200*
.973
31
.617
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 2: Leadership Styles and Respect, Normality Assumption
70
Independence.
In all independence plots (Figure 8 -13), it was observed that the values fell
within a band of -2 to 2 on the Y-axis with only one exception, which was always very
close to the upper bound line. The points did not follow a pattern (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Figure 8: Leadership Styles and Respect – Independence Plot 1
71
Figure 9: Leadership Styles and Respect – Independence Plot 2
Figure 10: Leadership Styles and Respect – Independence Plot 3
72
Figure 11: Leadership Styles and Respect – Independence Plot 4
Figure 12: Leadership Styles and Respect – Independence Plot 5
73
Figure 13: Leadership Styles and Respect – Independence Plot 6
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 8 – 13)
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 14) were produced between all independent and
dependent variables. Only observe linear or random patterns were observable. Verifying
above independence plots 2 to 6, plots 4 and 6 showed a slight power pattern. As the x
74
and y values were restricted to 0 to 4, and the shown pattern was slight, it was acceptable
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 14: Leadership Styles and Respect – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
75
Normality.
The observed values followed the line in the normal Q-Q plot (Figure 15) in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
Figure 15: Leadership Styles and Respect – Normality Q-Q Plot
The skewness and kurtosis statistics (Table 3) were within the range of absolute
value 2.0 suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.0414
95% Confidence Interval for
Mean
Lower Bound
-.084
Upper Bound
.084
5% Trimmed Mean
-.003
Median
-.038
Variance
.053
Std. Deviation
.230
Minimum
-.369
Maximum
.457
76
Range
.827
Interquartile Range
.411
Skewness
.221
.421
Kurtosis
-.989
.821
Table 3: Leadership Styles and Respect – Normality Descriptive Statistics
The histogram (Figure 16) of unstandardized residuals reflected an acceptable
normal distribution suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
Figure 16: Leadership Styles and Respect – Normality Histogram
Examination of the below boxplot (Figure 17) suggested evidence of normality
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
77
Figure 17: Leadership Styles and Respect – Normality Boxplot
Both tests of normality (Table 4) provided non-significant tests results supporting
the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.120
31
.200*
.962
31
.330
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 4: Leadership Styles and Respect – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinerity.
The VIF (Table 5) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Collinearity Statistics
Tolerance
VIF
1
(Constant)
CR
.152
6.590
MBEA
.576
1.736
78
MBEP
.562
1.781
LV
.368
2.719
TL
.145
6.917
Group
.816
1.226
Table 5: Leadership Styles and Proactivity – Multicollinearity VIF
Only the eigenvalue of TL (Table 6) was close to 0 as the Group variable is
nominal and not relevant. A regression between TL and the other independent variables
produced an R2 of .855 and thus below the threshold of .9. No multicollinearity was
supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
6.011
2
.653
3
.172
4
.079
5
.072
6
.008
7
.004
Table 6: Leadership Styles and Proactivity – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Leadership Styles and Proactivity.
Leadership Styles and Proactivity - Outlier Analysis.
The centered leverage value cut-off value in the following plot (Figure 18) was
0.46875 for the centered leverage value calculated as described in Chapter 3. Case 8
might be an outlier as in both criteria it is noticeable. The cases 7, 14, 15, 29, and 29
might be outliers due to Cook’s D.
79
Figure 18: Leadership Styles and Proactivity, Outlier Cook’s D / Centered Leverage
In the following plot (Figure 19) no outlier could be observed, all data points are
outside the 3 threshold.
Figure 19: Leadership Styles and Proactivity, Outlier Externally Studentized residuals
No case was identified three times as a potential outlier.
80
Leadership Styles and Proactivity - Assumption Analysis.
Normality of Proactivity.
Both tests of normality (Table 7) were not significant, and the assumption was
reasonably fulfilled (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
Df
Sig.
Statistic
Df
Sig.
Proactivity
.088
32
.200*
.972
32
.566
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 7: Leadership Style and Proactivity, Normality
Independence.
In all independence plots (figures 20 to 25), it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with only one exception, which was always
very close to the upper bound line. The points did not follow a pattern (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
81
Figure 20: Leadership Styles and Proactivity – Independence Plot 1
Figure 21: Leadership Styles and Proactivity – Independence Plot 2
Figure 22: Leadership Styles and Proactivity – Independence Plot 3
82
Figure 23: Leadership Styles and Proactivity – Independence Plot 4
Figure 24: Leadership Styles and Proactivity – Independence Plot 5
83
Figure 25: Leadership Styles and Proactivity – Independence Plot 6
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (figures 20 to 25)
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (figure 26) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
independence plots 2 to 6, plots 4 and 6 showed a slight power pattern. As the x and y
84
values were restricted to 0 to 4, and the shown pattern was slight, it was acceptable
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 26: Leadership Styles and Proactivity – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
85
Normality.
The observed values followed the line in the normal Q-Q plot of unstandardized
residuals (Figure 27) in an acceptable manner suggesting evidence of normality (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 27: Leadership Styles and Proactivity – Normality Q-Q Plot
The skewness and kurtosis statistics (Table 8) were within the range of absolute
value 2.0 suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.051
95% Confidence Interval for
Mean
Lower Bound
-.104
Upper Bound
.104
5% Trimmed Mean
-.004
Median
-.062
Variance
.083
Std. Deviation
.288
Minimum
-.491
Maximum
.561
86
Range
1.052
Interquartile Range
.415
Skewness
.376
.414
Kurtosis
-.600
.809
Table 8: Leadership Styles and Proactivity – Normality Descriptive Statistics
The histogram (Figure 28) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 28: Leadership Styles and Proactivity – Normality Histogram
Examination of the boxplot (Figure 29) suggested no challenge to evidence of
normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
87
Figure 29: Leadership Styles and Proactivity – Normality Boxplot
Both tests of normality (Table 9) provided non-significant tests results supporting
the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.141
32
.109
.962
32
.304
a. Lilliefors Significance Correction
Table 9: Leadership Styles and Proactivity – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 10) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tolerance
VIF
1
(Constant)
88
TL
.145
6.918
CR
.159
6.275
MBEA
.658
1.519
MBEP
.623
1.605
LV
.377
2.652
group
.867
1.154
Table 10: Leadership Styles and Proactivity – Multicollinearity VIF
Only the eigenvalue (Table 11) of Laissez Fair was close to 0 as the Group
variable was nominal and not relevant. A regression between Laissez Fair and the other
independent variables produced an R2 of .623 and thus below the threshold of .9. No
multicollinearity was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
6.014
2
.648
3
.171
4
.078
5
.075
6
.009
7
.004
Table 11: Leadership Styles and Proactivity – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Leadership Styles and Innovative Work Behavior.
Leadership Styles and Innovative Work Behavior - Outlier Analysis.
The centered leverage value (Figure 30) cut-off value was 0.46875 calculated as
described in Chapter 3. Case 8 might be an outlier. Even though case 11 was not critical
89
of the centered leverage value, an outlier possibility is strongly supported by Cook’s D
(Figure 30).
Figure 30: Leadership Styles and IWB, Outlier Cook’s D / Centered Leverage
Cases 11 (Figure 31) might be an outlier as it is located outside the threshold of 3
and close to 4.
Figure 31: Leadership Styles and IWB, Outlier Externally Studentized residuals
90
Cases 8 and 11 were excluded because in both cases two outlier verifications were
rather strong.
Leadership Styles and Innovative Work Behavior - Assumption Analysis.
Normality of Innovative Work Behavior.
No test of normality (Table 12) was significant. But one test was close to
significance, and further analysis was conducted.
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
Df
Sig.
Innovative Workbehavior
.122
30
.200*
.932
30
.056
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 12: Leadership Styles and IWB, Normality of IWB - Tests
The observed values in the normal Q-Q plot (Figure 31) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
91
Figure 32: Leadership Styles and IWB, Normality of IWB – Normality Q-Q Plot
The skewness and kurtosis statistics (Table 13) were within the range of absolute
value 2.0 suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Innovative
Workbehavior
Mean
2.909
.068
95% Confidence Interval for
Mean
Lower
Bound
2.770
Upper
Bound
3.048
5% Trimmed Mean
2.888
Median
2.859
Variance
.139
Std. Deviation
.372
Minimum
2.333
Maximum
4.000
Range
1.666
Interquartile Range
.638
Skewness
.900
.427
Kurtosis
.903
.833
Table 13: Leadership Styles and IWB, Normality of IWB - Descriptive
92
The histogram (Figure 33) reflected a still acceptable normal distribution
suggesting evidence of normality, but the one data point to the right as the positive
skewness are challenging (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 33: Leadership Styles and IWB, Normality of IWB – Normality Histogram
Examination of the above boxplot (Figure 34) suggested evidence of normality
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
93
Figure 34: Leadership Styles and IWB, Normality of IWB – Normality Boxplot
Considering all evidence from above, the normality assumption was reasonably
fulfilled (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 35 to 40), it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with only one data point exception. The points
did not follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
94
Figure 35: Leadership Styles and IWB – Independence Plot 1
Figure 36: Leadership Styles and IWB – Independence Plot 2
95
Figure 37: Leadership Styles and IWB – Independence Plot 3
Figure 38: Leadership Styles and IWB – Independence Plot 4
96
Figure 39: Leadership Styles and IWB – Independence Plot 5
Figure 40: Leadership Styles and IWB – Independence Plot 6
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
97
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 35 to
40) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 41) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
independence plots 2 to 6 linear patterns were supported (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
98
Figure 41: Leadership Styles and IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values in the normal Q-Q plot (Figure 42) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
99
Figure 42: Leadership Styles and IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 14) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.0423
95% Confidence Interval for
Mean
Lower Bound
-.086
Upper Bound
.086
5% Trimmed Mean
-.001
Median
.015
Variance
.054
Std. Deviation
.231
Minimum
-.584
Maximum
.627
Range
1.211
Interquartile Range
.274
Skewness
.139
.427
Kurtosis
1.377
.833
Table 14: Leadership Styles and IWB – Normality Descriptive Statistics
The histogram (Figure 43) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
100
Figure 43: Leadership Styles and IWB – Normality Histogram
Examination of the below boxplot (Figure 44) challenged the evidence of
normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013
Figure 44: Leadership Styles and IWB – Normality Boxplot
101
Both tests of normality (Table 15) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.097
30
.200*
.977
30
.731
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 15: Leadership Styles and IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled but challenged.
Multicollinearity.
The VIF (Table 16) was above 10 suggesting multicollinearity is present between
CR and TL. (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013). To overcome the
multicollinearity one variable needed to be dropped (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013). As the literature indicates that TL is a stronger influence factor on
positive employee behavior than CR, CR will be dropped in the next model.
Model
Tolerance
VIF
1
(Constant)
CR
.086
11.653
MBEA
.458
2.183
MBEP
.607
1.648
LF
.371
2.696
TL
.068
14.811
group
.836
1.197
Table 16: Leadership Styles and IWB – Multicollinearity VIF
102
Leadership Styles (no CR) and Innovative Work Behavior - Assumption Analysis.
Normality of Innovative Work Behavior.
Both tests of normality (Table 17) were not significant, and the assumption is
reasonably fulfilled (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
Df
Sig.
Statistic
df
Sig.
Proactivity
.088
32
.200*
.972
32
.566
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 17: Leadership Styles/CR and IWB – Normality IWB
Independence.
In all plots of independence (Figures 45 to 49), it could be observed that the
values fell within a band of -2 to 2 on the Y-axis with two exceptions. The points did not
follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 45: Leadership Styles/CR and IWB – Independence Plot 1
103
Figure 46: Leadership Styles/CR and IWB – Independence Plot 2
Figure 47: Leadership Styles/CR and IWB – Independence Plot 3
104
Figure 48: Leadership Styles/CR and IWB – Independence Plot 4
Figure 49: Leadership Styles/CR and IWB – Independence Plot 5
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
105
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 45 to
49) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 50) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
independence plots 2 to 5, the linearity assumption was acceptably fulfilled (Lomax &
Hahs-Vaughn, 2012; Cohen et al., 2013).
106
Figure 50: Leadership Styles/CR and IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values in the normal Q-Q plot (Figure 51) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
107
Figure 51: Leadership Styles/CR and IWB – Normality Q-Q Plot
The skewness and kurtosis statistics (Table 18) were within the range of absolute
value 2.0 suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.042
95% Confidence Interval for
Mean
Lower Bound
-.086
Upper Bound
.086
5% Trimmed Mean
-.001
Median
.020
Variance
.054
Std. Deviation
.232
Minimum
-.595
Maximum
.621
Range
1.216
Interquartile Range
.257
Skewness
.087
.427
Kurtosis
1.414
.833
Table 18: Leadership Styles/CR and IWB – Normality Descriptive Statistics
108
The histogram (Figure 52) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 52: Leadership Styles/CR and IWB – Normality Histogram
Examination of the boxplot (Figure 53) suggested a challenge to evidence of
normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
109
Figure 53: Leadership Styles/CR and IWB – Normality Boxplot
Both tests of normality (Table 19) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.100
30
.200*
.975
30
.688
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 19: Leadership Styles/CR and IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 20) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
110
Tolerance
VIF
1
(Constant)
MBEA
.533
1.877
MBEP
.607
1.648
LF
.371
2.695
group
.837
1.194
TL
.320
3.122
Table 20: Leadership Styles/CR and IWB – Multicollinearity VIF
Only the eigenvalue of TL (Table 21) was close to 0. A regression between TL
and the other independent variables produced an R2 of .68 and thus below the threshold
of .9. No multicollinearity was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al.,
2013).
Model
Dimension
Eigenvalue
1
1
5.127
2
.572
3
.174
4
.070
5
.051
6
.007
Table 21: Leadership Styles/CR and IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Mediation Model – TL through Respect to Proactivity.
TL through Respect to Proactivity - Outlier Analysis.
The centered leverage value (Figure 54) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 29 might be an outlier as in both criteria it is noticeable.
111
Figure 54: TL through Respect to Proactivity, Outlier Cook’s D / Centered Leverage
Cases 11 might be an outlier (Figure 55).
Figure 55: TL through Respect to Proactivity, Outlier Externally Studentized residuals
No case was identified three times as a potential outlier.
112
TL through Respect to Proactivity - Assumption Analysis.
Normality of Proactivity.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 56 to 58), it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with only one exception. The points did not
follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 56: TL through Respect to Proactivity – Independence Plot 1
113
Figure 57: TL through Respect to Proactivity – Independence Plot 2
Figure 58: TL through Respect to Proactivity – Independence Plot 3
114
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 56 to
58) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 59) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
independence plots 2 and 3 linear patterns were supported (Lomax & Hahs-Vaughn,
2012; Cohen et al., 2013).
115
Figure 59: TL through Respect to Proactivity – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values in the normal Q-Q plot (Figure 60) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
116
Figure 60: TL through Respect to Proactivity – Normality Q-Q Plot
The skewness and kurtosis (Table 22) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.052
95% Confidence Interval for
Mean
Lower Bound
-.106
Upper Bound
.106
5% Trimmed Mean
-.004
Median
.019
Variance
.088
Std. Deviation
.296
Minimum
-.568
Maximum
.762
Range
1.330
Interquartile Range
.438
Skewness
.201
.414
Kurtosis
.134
.809
Table 22: TL through Respect to Proactivity – Normality Descriptive Statistics
117
The histogram (Figure 61) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 61: TL through Respect to Proactivity – Normality Histogram
The boxplot (Figure 62) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
118
Figure 62: TL through Respect to Proactivity – Normality Boxplot
Both tests of normality (Table 23) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.080
32
.200*
.985
32
.924
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 23: TL through Respect to Proactivity – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 24) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Collinearity Statistics
Tolerance
VIF
119
1
(Constant)
TL
.293
3.411
Respect
.288
3.467
group
.969
1.032
Table 24: TL through Respect to Proactivity – Multicollinearity VIF
Only the eigenvalue (Table 25) of Group was close to 0, but the Group variable
was nominal and not relevant. No multicollinearity was supported (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Collinearity Diagnosticsa
Model
Dimension
Eigenvalue
Condition Index
Variance Proportions
(Constant)
TL
Respect
group
1
1
3.894
1.000
.00
.00
.00
.01
2
.081
6.936
.00
.03
.00
.85
3
.022
13.246
.32
.26
.00
.14
4
.003
34.825
.67
.71
.99
.01
a. Dependent Variable: Proactivity
Table 25: TL through Respect to Proactivity – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Mediation model – TL through Respect to Innovation Work Behavior.
TL through Respect to Innovation Work Behavior - Outlier Analysis.
The centered leverage value (Figure 63) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 29 might be an outlier. But also cases 7, 9, and 11 required
attention due to their Cook’s D values.
120
Figure 63: TL through Respect to IWB, Outlier Cook’s D / Centered Leverage
Cases 9 and 11 might be outliers (Figure 64).
Figure 64: TL through Respect to IWB, Outlier Externally Studentized residuals
121
Cases 9 and 11 were identified two times as potential outliers. Both cases were
excluded from the following mediation analysis.
TL through Respect to Innovation Work Behavior - Assumption Analysis.
Normality of Innovation Work Behavior.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 65 to 67), it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with two exceptions. The points did not follow
a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 65: TL through Respect to IWB – Independence Plot 1
122
Figure 66: TL through Respect to IWB – Independence Plot 2
Figure 67: TL through Respect to IWB – Independence Plot 3
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
123
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 65 to
67) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 68) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
independence plots 2 and 3 linear patterns were supported (Lomax & Hahs-Vaughn,
2012; Cohen et al., 2013).
124
Figure 68: TL through Respect to IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values of the normal Q-Q plot (Figure 69) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
125
Figure 69: TL through Respect to IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 26) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.036
95% Confidence Interval for
Mean
Lower Bound
-.073
Upper Bound
.073
5% Trimmed Mean
.001
Median
-.012
Variance
.039
Std. Deviation
.197
Minimum
-.487
Maximum
.431
Range
.918
Interquartile Range
.295
Skewness
-.023
.427
Kurtosis
.256
.833
Table 26: TL through Respect to IWB – Normality Descriptive Statistics
126
The histogram (Figure 70) reflected a challenge to the normal distribution (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 70: TL through Respect to IWB – Normality Histogram
The boxplot (Figure 71) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
127
Figure 71: TL through Respect to IWB – Normality Boxplot
Both tests of normality (Table 27) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.092
30
.200*
.981
30
.851
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 27: TL through Respect to IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 28) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
128
Tolerance
VIF
1
(Constant)
Respect
.307
3.253
TL
.314
3.181
group
.961
1.041
Table 28: TL through Respect to IWB – Multicollinearity VIF
Only the eigenvalue (Table 29) of Group was close to 0, but the Group variable
was nominal and not relevant. No multicollinearity was supported (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
3.893
2
.081
3
.023
4
.003
Table 29: TL through Respect to IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Double Mediation model – TL through Respect and Proactivity to
Innovation Work Behavior.
TL through Respect and Proactivity to Innovation Work Behavior - Outlier Analysis.
The centered leverage value (Figure 72) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 29 might be outlier. But cases 7 and 11 required attention
due to their Cook’s D values.
129
Figure 72: TL through Respect and Proactivity to IWB, Outlier Cook’s D / Centered
Leverage
Cases 7 and 11 might be outliers (Figure 73).
Figure 73: TL through Respect and Proactivity to IWB, Outlier Externally Studentized
residuals
130
Cases 7 and 11 were identified two times as potential outliers. Both cases were
excluded from the following mediation analysis.
TL through Respect and Proactivity to Innovation Work Behavior – Assumption
Analysis.
Normality of IWB.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figure 74 to 77) it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with two exceptions not outside of absolute
value 3. The points did not follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al.,
2013).
Figure 74: TL through Respect and Proactivity to IWB – Independence Plot 1
131
Figure 75: TL through Respect and Proactivity to IWB – Independence Plot 2
Figure 76: TL through Respect and Proactivity to IWB – Independence Plot 3
132
Figure 77: TL through Respect and Proactivity to IWB – Independence Plot 4
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 74 to
77) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 78) were produced between all independent and
dependent variables. Only linear or random patterns were observable. Verifying above
133
independence plots, linear patterns were supported (Figures 75 to 77) (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Figure 78: TL through Respect and Proactivity to IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values in the Normal Q-Q plot (Figure 79) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
134
Figure 79: TL through Respect and Proactivity to IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 30) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.032
95% Confidence Interval for
Mean
Lower Bound
-.065
Upper Bound
.065
5% Trimmed Mean
-.001
Median
-.018
Variance
.031
Std. Deviation
.175
Minimum
-.390
Maximum
.449
Range
.840
Interquartile Range
.194
Skewness
.272
.427
Kurtosis
.787
.833
Table 30: TL through Respect and Proactivity to IWB – Normality Descriptive Statistics
135
The histogram (Figure 80) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 80: TL through Respect and Proactivity to IWB – Normality Histogram
The boxplot (Figure 81) reflected a challenge to normal distribution (Lomax &
Hahs-Vaughn, 2012; Cohen et al., 2013).
136
Figure 81: TL through Respect and Proactivity to IWB – Normality Boxplot
Both tests of normality (Table 31) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.091
30
.200*
.981
30
.863
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 31: TL through Respect and Proactivity to IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 32) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012 and Cohen, Cohen, West &Aiken, 2013).
Model
Tolerance
VIF
137
1
(Constant)
TL
.334
2.990
Respect
.288
3.469
Proactivity
.495
2.021
group
.833
1.200
Table 32: TL through Respect and Proactivity to IWB – Multicollinearity VIF
Only the eigenvalue (Table 33) of Proactivity was close to 0. The Group variable
was nominal and not relevant. A regression between Proactivity and the other
independent variables produced an R2 of .505 and thus below the threshold of .9. No
multicollinearity was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
4.886
2
.084
3
.021
4
.006
5
.003
Table 33: TL through Respect and Proactivity to IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
138
Double Mediation model – CR through Respect and Proactivity to
Innovation Work Behavior.
CR through Respect and Proactivity to Innovation Work Behavior - Outlier Analysis.
The centered leverage value (Figure 82) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 8 and 29 might be an outlier. But cases 7 and 11 require
attention due to their Cook’s D values.
Figure 82: CR through Respect and Proactivity to IWB, Outlier Cook’s D / Centered
Leverage
Cases 7 and 11 might be outliers (Figure 83).
139
Figure 83: CR through Respect and Proactivity to IWB, Outlier Externally Studentized
residuals
Cases 7 and 11 were identified two times as potential outliers. Both cases were
excluded from the following mediation analysis.
CR through Respect and Proactivity to Innovation Work Behavior - Assumption
Analysis.
Normality of Innovation Work Behavior.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 84 to 87), it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with one exception, still inside absolute value
3. The points did not follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al.,
2013).
140
Figure 84: CR through Respect and Proactivity to IWB – Independence Plot 1
Figure 85: CR through Respect and Proactivity to IWB – Independence Plot 2
141
Figure 86: CR through Respect and Proactivity to IWB – Independence Plot 3
Figure 87: CR through Respect and Proactivity to IWB – Independence Plot 4
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
142
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 84 to
87) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
linearity.
Multiple scatterplots (Figure 88) were produced between all independent and
dependent variables. Only linear and random patterns were observable. Verifying above
independence plots, linear patterns were supported (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
143
Figure 88: CR through Respect and Proactivity to IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
Normality.
The observed values in the normal Q-Q plot (Figure 89) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
144
Figure 89: CR through Respect and Proactivity to IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 34) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.032
95% Confidence Interval for
Mean
Lower Bound
-.066
Upper Bound
.066
5% Trimmed Mean
-.003
Median
-.028
Variance
.032
Std. Deviation
.178
Minimum
-.362
Maximum
.458
Range
.821
Interquartile Range
.218
Skewness
.342
.427
Kurtosis
.400
.833
Table 34: CR through Respect and Proactivity to IWB – Normality Descriptive Statistics
145
The histogram (Figure 90) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 90: CR through Respect and Proactivity to IWB – Normality Histogram
The box plot (Figure 91) reflected a challenge to the normal distribution (Lomax
& Hahs-Vaughn, 2012; Cohen et al., 2013).
146
Figure 91: CR through Respect and Proactivity to IWB – Normality Boxplot
Both tests of normality (Table 35) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.110
30
.200*
.985
30
.945
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 35: CR through Respect and Proactivity to IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 36) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
147
Tolerance
VIF
1
(Constant)
CR
.484
2.068
Respect
.391
2.560
Proactivity
.491
2.039
group
.834
1.199
Table 36: CR through Respect and Proactivity to IWB – Multicollinearity VIF
Only the eigenvalue (Table 37) of Proactivity was close to 0. The Group variable
is nominal and not relevant. A regression between Proactivity and the other independent
variables produced an R2 of .509 and thus below the threshold of .9. No multicollinearity
was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
4.885
2
.084
3
.021
4
.007
5
.004
Table 37: CR through Respect and Proactivity to IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Double Mediation Model – MBEA through Respect and Proactivity to
Innovation Work Behavior.
MBEA through Respect and Proactivity to Innovation Work Behavior - Outlier
Analysis.
The centered leverage value (Figure 92) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 8 and 18 might be an outlier. But cases 7 and 11 required
attention due to their Cook’s D values.
148
Figure 92: MBEA through Respect and Proactivity to IWB, Outlier Cook’s D /
Centered Leverage
Cases 7 and 11 might be outliers (Figure 93).
149
Figure 93: MBEA through Respect and Proactivity to IWB, Outlier Externally
Studentized residuals
Cases 7 and 11 were identified two times as potential outliers. Both cases were
excluded from the following mediation analysis.
MBEA through Respect and Proactivity to Innovation Work Behavior - Assumption
Analysis.
Normality of Innovation Work Behavior.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 94 to 97) it could be observed that the values
fell within a band of -2 to 2 on the Y-axis with one exception. The points did not follow a
pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 94: MBEA through Respect and Proactivity to IWB – Independence Plot 1
150
Figure 95: MBEA through Respect and Proactivity to IWB – Independence Plot 2
Figure 96: MBEA through Respect and Proactivity to IWB – Independence Plot 3
151
Figure 97: MBEA through Respect and Proactivity to IWB – Independence Plot 4
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 94 to
97) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 98) were produced between all independent and
dependent variables. Only linear and random patterns were observable. Verifying above
152
independence plots, linear patterns were supported (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
Figure 98: MBEA through Respect and Proactivity to IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
153
Normality.
The observed values in the normal Q-Q Plot (Figure 99) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
Figure 99: MBEA through Respect and Proactivity to IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 38) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.032
95% Confidence Interval for
Mean
Lower Bound
-.066
Upper Bound
.066
5% Trimmed Mean
-.001
Median
-.032
Variance
.032
154
Std. Deviation
.178
Minimum
-.399
Maximum
.449
Range
.849
Interquartile Range
.226
Skewness
.214
.427
Kurtosis
.585
.833
Table 38: MBEA through Respect and Proactivity to IWB – Normality Descriptive
Statistics
The histogram (Figure 100) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 100: MBEA through Respect and Proactivity to IWB – Normality Histogram
The box plot (Figure 101) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
155
Figure 101: MBEA through Respect and Proactivity to IWB – Normality Boxplot
Both tests of normality (Table 39) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.102
30
.200*
.986
30
.948
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 39: MBEA through Respect and Proactivity to IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinerity.
The VIF (Table 40) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Collinearity Statistics
156
Tolerance
VIF
1
(Constant)
MBEA
.813
1.230
Respect
.559
1.790
Proactivity
.428
2.334
group
.850
1.177
Table 40: MBEA through Respect and Proactivity to IWB – Multicollinearity VIF
Only the eigenvalue (Table 41) of Proactivity was close to 0. The Group variable
is nominal and not relevant. A regression between Proactivity and the other independent
variables produced an R2 of .572 and thus below the threshold of .9. No multicollinearity
was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
4.821
2
.100
3
.066
4
.008
5
.004
Table 41: MBEA through Respect and Proactivity to IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Double Mediation model – MBEP through Respect and Proactivity to
Innovation Work Behavior.
MBEP through Respect and Proactivity to Innovation Work Behavior – Outlier
Analysis.
The centered leverage value (Figure 102) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 7 and 20 might be outliers. But case 11 required attention
due to its Cook’s D value.
157
Figure 102: MBEP through Respect and Proactivity to IWB, Outlier Cook’s D /
Centered Leverage
Cases 7 and 11 might be outliers (Figure 103).
158
Figure 103: MBEP through Respect and Proactivity to IWB, Outlier Externally
Studentized residuals
Cases 7 was identified three times as potential outliers and was excluded from the
following mediation analysis.
MBEP through Respect and Proactivity to Innovation Work Behavior - Assumption
Analysis.
Normality of Innovation Work Behavior.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 104 to 107), it could be observed that the
values fell within a band of -2 to 2 on the Y-axis with one exception. The points did not
follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
159
Figure 104: MBEP through Respect and Proactivity to IWB – Independence Plot 1
Figure 105: MBEP through Respect and Proactivity to IWB – Independence Plot 2
Figure 106: MBEP through Respect and Proactivity to IWB – Independence Plot 3
160
Figure 107: MBEP through Respect and Proactivity to IWB – Independence Plot 4
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 104 to
107) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 108) were produced between all independent and
dependent variables. Only linear and random patterns were observable. Verifying above
161
independence plots, linear patterns were supported (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
Figure 108: MBEP through Respect and Proactivity to IWB – Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
162
Normality.
The observed values in the normal Q-Q plot (Figure 109) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
Figure 109: MBEP through Respect and Proactivity to IWB – Normality Q-Q Plot
The skewness and kurtosis (Table 42) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.037
95% Confidence Interval for
Mean
Lower Bound
-.076
Upper Bound
.076
5% Trimmed Mean
-.007
Median
-.005
Variance
.043
Std. Deviation
.207
Minimum
-.332
163
Maximum
.514
Range
.847
Interquartile Range
.296
Skewness
.342
.421
Kurtosis
-.039
.821
Table 42: MBEP through Respect and Proactivity to IWB – Normality Descriptive
Statistics
The histogram (Figure 109) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
Figure 110: MBEP through Respect and Proactivity to IWB – Normality Histogram
The box plot (Figure 111) reflected an acceptable normal distribution suggesting
evidence of normality (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013.
164
Figure 111: MBEP through Respect and Proactivity to IWB – Normality Boxplot
Both tests of normality (Table 43) provided non-significant tests results
supporting the normality assumption (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Tests of Normality
Kolmogorov-Smirnova
Shapiro-Wilk
Statistic
df
Sig.
Statistic
df
Sig.
Unstandardized Residual
.110
31
.200*
.969
31
.488
*. This is a lower bound of the true significance.
a. Lilliefors Significance Correction
Table 43: MBEP through Respect and Proactivity to IWB – Normality Test
Considering all evidence from above, the normality assumption was reasonably
fulfilled.
Multicollinearity.
The VIF (Table 44) was below 10 suggesting no multicollinearity was present
(Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Collinearity Statistics
165
Tolerance
VIF
1
(Constant)
MBEP
.923
1.083
Respect
.598
1.672
Proactivity
.557
1.796
group
.851
1.175
Table 44: MBEP through Respect and Proactivity to IWB – Multicollinearity VIF
Only the eigenvalue (Table 45) of Proactivity was close to 0. The Group variable
was nominal and not relevant. A regression between Proactivity and the other
independent variables produced an R2 of .443 and thus below the threshold of .9. No
multicollinearity was supported (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Model
Dimension
Eigenvalue
1
1
4.696
2
.229
3
.061
4
.008
5
.005
Table 45: MBEP through Respect and Proactivity to IWB – Multicollinearity Eigenvalue
Considering all evidence from above, the no multicollinearity assumption was
reasonably fulfilled.
Double Mediation model – LF through Respect and Proactivity to Innovation
Work Behavior.
LF through Respect and Proactivity to Innovation Work Behavior - Outlier Analysis.
The centered leverage value (Figure 112) cut-off value was 0.28125 calculated as
described in Chapter 3. Case 7 might be an outlier. But case 11 required attention due to
its Cook’s D value.
166
Figure 112: LF through Respect and Proactivity to IWB, Outlier Cook’s D / Centered
Leverage
Cases 7 and 11 might be outliers (Figure 113).
Figure 113: LF through Respect and Proactivity to IWB, Outlier Externally
Studentized residuals
167
Cases 7 was identified three times as potential outliers and was excluded from the
following mediation analysis.
LF through Respect and Proactivity to Innovation Work Behavior - Assumption
Analysis.
Normality of Innovation Work Behavior.
The normality assumption was reasonably fulfilled earlier (Lomax & Hahs-
Vaughn, 2012; Cohen et al., 2013).
Independence.
In all independence plots (Figures 114 to 117), it could be observed that the
values fell within a band of -2 to 2 on the Y-axis with one exception. The points did not
follow a pattern (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Figure 114: LF through Respect and Proactivity to IWB – Independence Plot 1
168
Figure 115: LF through Respect and Proactivity to IWB – Independence Plot 2
Figure 116: LF through Respect and Proactivity to IWB – Independence Plot 3
169
Figure 117: LF through Respect and Proactivity to IWB – Independence Plot 4
Considering all evidence from above, the independence assumption was
reasonably fulfilled.
Homogeneity of Variance.
The spreads of the studentized residuals over the range of unstandardized
predicted values and over all independent variables were fairly constant (Figures 114 to
117) (Lomax & Hahs-Vaughn, 2012; Cohen et al., 2013).
Considering this evidence, the homoscedasticity assumption was reasonably
fulfilled.
Linearity.
Multiple scatterplots (Figure 118) were produced between all independent and
dependent variables. Only linear and random patterns were observable. Verifying above
170
independence plots, linear patterns were supported (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013).
Figure 118: LF through Respect and Proactivity to Innovation Work Behavior –
Linearity
Considering all evidence from above, the linearity assumption was reasonably
fulfilled.
171
Normality.
The observed values in the normal Q-Q plot (Figure 119) followed the line in an
acceptable manner suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012;
Cohen et al., 2013.
Figure 119: LF through Respect and Proactivity to Innovation Work Behavior –
Normality Q-Q Plot
The skewness and kurtosis (Table 46) were within the range of absolute value 2.0
suggesting evidence of normality (Lomax & Hahs-Vaughn, 2012).
Descriptive Statistic
Statistic
Std. Error
Unstandardized Residual
Mean
.000
.037
95% Confidence Interval for
Mean
Lower Bound
-.075
Upper Bound
.075
5% Trimmed Mean
-.005
Median
-.022
Variance
.043
Std. Deviation
.206
