Smart-PLS 4

Workshop
Theory and Practice

By
Dalowar Hossan
dalowarhossan.bd@gmail.com

1
 Research Title

Employee satisfaction mediates the relationship between

motivational factors and organizational performance; with leadership
style as a moderator

2
Research Framework & Hypothesis

3
Research Hypothesis

4
Use PLS-SEM when

✓ The goal is predicting key target constructs or identifying key “driver” constructs.
✓ Formatively measured constructs are part of the structural model. Note that
formative measures can also be used with CB-SEM, but doing so requires to
construct specification modifications (e.g., the construct must include both
formative and reflective indicators to meet identification requirements).
✓ The structural model is complex (many constructs and many indicators).
✓ The sample size is small and/or the data are nonnormally distributed.
✓ The plan is to use latent variable scores in subsequent analyses.

Use CB-SEM when (AMOS)

o The goal is theory testing, theory confirmation, or the comparison of alternative

theories.
o Error terms require additional specifications, such as covariation.
o The structural model has circular relationships.
o The research requires a global goodness-of-fit criterion.

5
THEORY

6
Fundamentals of PLS-SEM

7
Fundamentals of PLS-SEM

dalowarhossan.bd@gmail.com

8
Fundamentals of PLS-SEM
(Latent & Observed Variables)

dalowarhossan.bd@gmail.com

9
Formative Measurement Model

dalowarhossan.bd@gmail.com

10
Reflective Measurement Model

11
12
Reflective VS Formative Measurement Model

13
Formative VS Reflective Measurement Model

14
Formative VS Reflective Measurement Model

15
Structural Model

16
Key Considerations Prior to Data Analysis

17
Key Considerations Prior to Data Analysis
Sampling Technique

18
Key Considerations Prior to Data Analysis
Sampling Technique

19
Key Considerations Prior to Data Analysis
(Sample Size)

20
Key Considerations Prior to Data Analysis
Pre-test & Pilot-test

21
Key Considerations Prior to Data Analysis
Common Method Variance

22
Key Considerations Prior to Data Analysis
Reverse Coding

23
Key Considerations Prior to Data Analysis
EFA & CFA

dalowarhossan.bd@gmail.com

24
Key Considerations Prior to Data Analysis
Missing Value Imputation

25
Basic PLS Modelling

26
Measurement Model Assessment

A step by step approach

27
Assessment of Measurement Model

28
Reflective Model Assessment

29
Reflective Model Assessment

30
Reflective Model Assessment

dalowarhossan.bd@gmail.com

31
Convergent Validity: Factor Loadings

32
Convergent Validity: AVE

33
Reflective Model Assessment

34
Discriminant Validity: Fornell & Larcker (F&L)

35
Discriminant Validity: Fornell & Larcker (F&L)

36
Discriminant Validity: Fornell & Larcker (F&L)

37
Discriminant Validity: HTMT

38
Discriminant Validity: HTMT Ratio

39
Reporting HTMT Output Ratio

40
Higher Order Constructs

41
Higher Order Constructs

42
Higher Order Constructs

43
Mediation Analysis

44
Mediation Analysis

45
Moderation

46
Moderation

47
Moderation

48
 Preparation
i) Missing values in SPSS need to be addressed using EM.
ii) Save SPSS file as csv, SPSS, XL format.
iii) Remove variables/data which are not needed in file.

49
Getting Started with PLS

i) Create new project

ii) Double-click to import data.
iii) Draw diagram on workplace.

50
Path Modelling (Reflective indicators)

Assessment of measurement model (PLS algorithm)

a. Convergent validity (Outer loadings, AVE and CR)

b. Discriminant validity (Fornell & Larcker, Cross Loading and HTMT)

51
Path Modelling (Reflective indicators

Assessment of structural model

a. Assessment of collinearity (VIF) (PLS-SEM algorithm)

b. Assessment of path coefficient (Bootstrapping)
c. Coefficient of Determination (R square) (PLS-SEM algorithm)
d. Effect size (f square) (PLS-SEM algorithm)
e. Predictive relevance (Q square) (PLSpredict)

52
 Mediation Analysis
i) Assessment of measurement model (PLS algorithm)

a. Convergent validity (Outer loadings, AVE and CR)

b. Discriminant validity (Fornell & Larcker and HTMT)

ii) Assessment of structural model

a. Assessment of collinearity (VIF) (PLS-SEM algorithm)
b. Assessment of path coefficient (Bootstrapping)
c. Assessment of indirect effect: (Bootstrapping 2-tailed)
d. Coefficient of Determination (R square) (PLS algorithm)
e. Effect size (f square) (PLS-SEM algorithm)
f. Predictive relevance (Q square) (PLSpredict)

53
 Moderator Analysis
(categorical data, which requires dummy coding)

i) Draw diagram, include moderator as IV

ii) Assessment of measurement model (PLS algorithm)

a. Convergent validity (Outer loadings, AVE and CR)
b. Discriminant validity (Fornell & Larcker and HTMT)
c. Assessment of R square and f square before moderation

54
 Moderator Analysis
(categorical data, which requires dummy coding)

iii) Add moderating effect, identify moderator and IV, select product
indicator and unstandardized product term

iv) Assessment of structural model

a. Assessment of collinearity (VIF) (PLS-SEM algorithm)
b. Assessment of path coefficient (Bootstrapping)
c. Assessment of path coefficient of moderation (Bootstrapping)
d. Coefficient of Determination (R square) (PLS-SEM algorithm)
e. Effect size (f square) (PLS-SEM algorithm)
55
 Moderator Analysis
(continuous data)
i) Draw diagram, include moderator as IV
ii) Assessment of measurement model (PLS algorithm)
a. Convergent validity (Outer loadings, AVE and CR)
b. Discriminant validity (Fornell & Larcker and HTMT)
c. Assessment of R square and f square before moderation
iii) Add moderating effect, identify moderator and IV, select product indicator and mean
centered product term
iv) Assessment of structural model
a. Assessment of collinearity (VIF) (PLS-SEM algorithm)
b. Assessment of path coefficient (Bootstrapping)
c. Assessment of path coefficient of moderation (Bootstrapping)
d. Coefficient of Determination (R square) (PLS-SEM algorithm)
e. Effect size (f square) (PLS-SEM algorithm)

56
 Reliability
Name of Index Level of Acceptance Literature Support

Cronbach Alpha > 0.70 or, 0.6 Robinson, Shaver & Wrightsman (1991)

Internal Consistency
Reliability

Composite Reliability > 0.70, Hair et al (2010), Hair et al (2014)

In exploratory research, 0.6 to 0.7 is
acceptable

57
 Convergent Validity
Name of Index Level of Acceptance Literature Support

Average Variance Explained AVE score > 0.5 Hair et al (2010), Hair et al
(AVE) (2014), Fornell and Larcker, 1981

Loadings for indicators > 0.50, Hair et al. (2007), Hair et al (2014)
0.4 to 0.7 may also be retained (if
Factor Loadings after removing indicator every time
the AVE and composite reliability
does not increase
58
 Discriminant Validity
Name of Index Level of Acceptance Literature Support

Fornell and Larcker criterion The square root of AVE of each construct Fornell and Larcker criterion (1981)
should be higher than its highest
correlation with any other constructs

Cross-Loadings Assessment An indicators' outer loadings

on a construct should be higher
than all its cross loadings with other
constructs

HTMT<.85 HTMT.85 – Kline (2011)

HTMT criterion
HTMT.90 – Gold, Malhotra, Segar
HTMT<.90 (2001) 59
Structural Model Assessment

60
Indices Structural Model
1. Path Co-efficient
2. R2
3. F2
4. Collinearity
5. Q2

61
 Indices for Structural Model
Path Co-efficient

Name of Index Level of Acceptance Literature Support

p value < 0.05 Hair et al., (2014)

Path Co-efficient
t value > 1.96 Hair et al., (2014)

62
 Indices for Structural Model
R2

Name of Index Level of Acceptance Literature Support

0.75 – Substantial

Co-efficient of determination 0.50 – Moderate Hair et al., (2014)

0.25 - Weak

63
 Indices for Structural Model
f2 ( f Square)

Name of Index Level of Acceptance Literature Support

0.35 – Large effect size

Hair et al., (2014)
Effect size to f2 0.15 – Medium effect size
Cohen (1988)
0.02 – Small effect size

64
 Indices for Structural Model
Colliniarity

Name of Index Level of Acceptance Literature Support

VIF < 3.3 Diamantopoulos & Sigouw (2006),

O’Brien (2007)
Variance Inflator Factor (VIF)

VIF < 5.0 Rogerson (2001)

65
 Indices for Structural Model
Q2 (Q Square)

Name of Index Level of Acceptance Literature Support

dalowarhossan.bd@gmail.com

Value larger than 0 indicates that Stone (1974)

Stone-Geisser Q2 Predictive exogenous constructs have Geisser (1974)
relevance predictive relevance over Hair et al., (2014)
endogenous construct

66
ASSESSMENT OF INTERNAL CONSISTENCY AND CONVERGENT VALIDITY

FROM XL SHEET

67
ASSESSMENT OF DISCRIMINANT VALIDITY
FROM XL SHEET

68
ASSESSMENT OF DISCRIMINANT VALIDITY
FROM XL SHEET

69
ASSESSMENT OF DISCRIMINANT VALIDITY
FROM XL SHEET

70
ASSESSMENT OF STRUCTURAL MODEL ANALYSIS
FROM XL SHEET

71
ASSESSMENT OF STRUCTURAL MODEL ANALYSIS
FROM XL SHEET

72
ASSESSMENT OF STRUCTURAL MODEL ANALYSIS
FROM XL SHEET

73
PRACTICE

74
Data Outliers

75
76
77
78
79
80
81
82
83
84
85
86
Z score ± 3.29

87
88
89
90
91
92
93
 Structural equation modeling is a multivariate statistical analysis technique that is used to analyze structural
relationships. This technique is a combination of factor analysis and multiple regression analysis, and it is used to
analyze the structural relationship between measured variables and latent constructs. This method is preferred by the
researcher because it estimates the multiple and interrelated dependence in a single analysis. Structural equation
modeling (SEM) is a powerful statistical technique that establishes measurement models and structural models. On the
other hand, multiple regression (MR) is considered a sophisticated and well-developed modeling approach to data
analysis with a history of more than 100 years.

94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
Inserted from SmartPLS 3

131
132
133
134
135
136
137
138
139
140
141
142
143
144
PLS-SEM MULTI-GROUP ANALYSIS

145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
 A regression model determines a
relationship between an independent
variable and a dependent variable, by
providing a function. Formulating a
regression analysis helps us to predict
the effects of the independent variable
on the dependent one.

The objective of multiple regression analysis is to use the independent variables whose values are known to
predict the single dependent value selected by the researcher. Each independent variable is weighted by the
regression analysis procedure to ensure maximal prediction from the set of independent variables.

There are three different types of regression models:

I. Linear
II. Non-linear
III. Multiple

160
A linear regression model is used to depict a relationship between variables that are proportional to
each other. Meaning, that the dependent variable increases/decreases with the independent variable.
The graphical representation has a straight linear line plotted between the variables. Even if the points
are not exactly in a straight line (which is always the case) we can still see a pattern and make sense of
it.

161
In the non-linear regression model, the graph doesn’t show a linear progression. Depending on how the
response variable reacts to the input variable, the line will rise or fall showing the height or depth of the
effect of the response variable.

To know that a non-linear regression model is the best fit for your scenario, make sure you look into your
variables and their patterns. If you see that the response variable is showing not-so-constant output to the
input variable, you can choose to use a non-linear model for your problem.

162
A multiple regression model is used when there is more than one independent variable affecting a
dependent variable. While predicting the outcome variable, it is important to measure how each of the
independent variables moves in their environment and how their changes will affect the output or target
variable.

163
164
165
166
167
168
 If the p-value is less than 0.05 and the t-value is higher than
1.96 then the effect is significant at confidence level of 95%.

Unstandardized beta represents the slope of the line between the predictor variable and the dependent variable. So for Variable 1,
every one unit increase in Variable LS1, the dependent variable decreases by 0.044 units. A standardized beta coefficient compares
the strength of the effect of each individual independent variable to the dependent variable. The higher the absolute value of the beta
coefficient, the stronger the effect. The range from 0 to 1 or 0 to -1, depending on the direction of the relationship.
169
The ANOVA tells us whether our regression model explains a statistically significant proportion of the variance. F and P say
“do the independent variables reliably predict the dependent variable?”
P value < alpha value (typically 0.05)

170
171
172
The coefficient of determination (R square) 0.556 suggests that 55.6% variance in the dependent variable
can be explained by independent variables.
1.5<d<2.5 (no first-order linear auto-correlation in the multiple linear regression data)

173
 VIF < 10

When IVs are correlated, there are problems in estimating regression coefficients. Collinearity
means that within the set of IVs, some of the IVs are (nearly) totally predicted by the other IVs.
174
 If one or more of the
eigenvalues are small
(close to zero) and the
corresponding condition
number is large,
then we have an
indication of
multicollinearity.

How to Deal with Collinearity (we have many options for it for example, )
➢ Factor analyze IVs to find sets of relatively homogeneous IVs that we can combine (add together).
➢ Use another type of analysis (path analysis, SEM).
175
For normally distributed data, observations should lie approximately in a straight line. If the data is non-normal, the
points form a curve

176
All residue values are arranged according to the frequency of normal distribution

177
Autocorrelation occurs when the residuals are not independent of each other. A random pattern of residuals
indicates the non-presence of autocorrelation

178
179
The PROCESS macro is essentially a modification to statistical programs that computes regression analyses containing
various combinations of mediators, moderators, and covariates.

180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201