Scribd
Upload
7 views

Using SmartPLS Software

The document provides an overview of Structural Equation Modeling (SEM) using SmartPLS software, detailing its history, types, and applications in social science research. It explains the differences between first and second generation statistical techniques, the components of path models, and criteria for evaluating PLS-SEM results. Additionally, it covers mediation analysis and practical considerations for using SEM effectively.

Uploaded by

seewendong102
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
7 views

Using SmartPLS Software

The document provides an overview of Structural Equation Modeling (SEM) using SmartPLS software, detailing its history, types, and applications in social science research. It explains the differences between first and second generation statistical techniques, the components of path models, and criteria for evaluating PLS-SEM results. Additionally, it covers mediation analysis and practical considerations for using SEM effectively.

Uploaded by

seewendong102
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 15

Using SmartPLS Software:

Structural Equation Modeling (PLS-


SEM)
Contents

• History of Structural • Path Models with • Assessing PLS-SEM • Evaluation of


Equation Model Structural Model
Latent Variables Results
• What is Structural
Equation Model? • Rule of Thumb for • Evaluation of • Mediation
• Consideration in Choosing between Measurement Model Analysis
Using SEM PLS-SEM and CB- • Practical
SEM Demonstration
• Social science researchers have been using
statistical tools for many years
• Application of first generation statistical History
methods e.g. factor analysis, regression
analysis was prevalent till 1980s
of SEM
• Second generation methods emerged since
1990s
• In some disciplines these represent almost
more than 50% in empirical research
• The emerging second generation tool is Partial
Least Squares Structural Equation Modeling
(PLS-SEM)
Researchers initially relied on Univariate and
Bivariate analysis to understand data and
relationships

SEM is used for more complex relationships


What is
SEM? SEM runs more sophisticated multivariate data
analysis methods

SEM involves the application of statistical


methods that simultaneously analyse multiple
variables
What is SEM? (Contd…)
Primarily Exploratory Primarily Confirmatory

First Generation • Cluster Analysis • Analysis of Variance


Technique • Exploratory Factor • Logistic Regression
Analysis • Multiple Regression
• Multidimensional • Confirmatory Factor
Scaling Analysis

Second Generation • Partial Least Squares • Co-variance Based


Technique Structural Equation Structural Equation
Modeling (PLS-SEM) Modeling (CB-SEM)
Consideration in Using SEM

1 2 3 4 5
Composite Measurement Measurement Coding Data
variables scales distribution
Path Models with
Latent Variables
• Exogenous Variable
• Whose value is independent from other
variables
• Endogenous Variable
• Whose value is dependent on other
variables
• Measurement Model/Outer Model
• Relationship between manifest and latent
variables
• Structural Model/Inner Model
• Relationship between exogenous and
endogenous latent variables
• Reflective Model
• Indicators are caused by the latent
variables (arrows are pointing from latent
to manifest variable)
• Formative Model
• Indicators cause the construct (arrows are
pointing from manifest to latent variable)
• The goal is predicting key target constructs

Use PLS- • Formatively measured constructs are part of the


structural model

SEM
• The structural model is complex
• The sample size is small
• 10 times of the indicators of the construct

Rule of Thumb when • 10 times of the structural paths


• Data is nonnormally distributed

for Choosing
between PLS-
SEM and CB-
SEM
Use CB- • The goal is theory testing

SEM
• Error terms require additional specification
• The research requires a global goodness-of-fit (GoF)
criterion

when
n of d e l Internal consistency
Convergent validity Discriminant validity
o
atio t M (Cronbach’s alpha,
(indicator reliability, (Fornell Larcker,
v al u m e n average variance Cross-Loadings,
E re composite reliability)
e a su extracted) HTMT Ratio)
M
Assessing
PLS-SEM ua
n of el
tio Mod
Assess structural
model for collinearity
Assess structural
Assess the Level of R2
al a l
Results
paths
Ev ctur issue
u
Str

Assess the effect size Predictive relevance


f2 q2
Evaluation of Measurement Model

1 2 3

Internal Consistency Convergent Validity Discriminant Validity


• Cronbach’s alpha (sensitive to • Indicator reliability (outer loading) • Fornell Larcker
number of items in the scale) • Latent variable should explain at least 50% • Square Root of AVE
• Composite reliability of the indicator’s variance i.e. 0.7082 • Should be greater than its highest
• Values between 0.6 and 0.7 • Items should be removed below 0.708 if correlation with any other construct
acceptable in exploratory research there is a problem with internal • Cross Loadings
consistency • Indicators outer loading should be
• Minimum acceptable value = 0.40 greater than any of its cross-loadings
• Values between 0.7 and 0.9
• AVE • HTMT Ratio
satisfactory • Value should be higher than 0.5 • Maximum threshold level is 0.90
• Values above 0.95 are less
• Minimum acceptable value = 0.4
desirable
Collinearity Assessment

• Variance inflation factor (VIF)


• Values < 5

Evaluation Structural Model Path Coefficients

of • Standardized values between -1 and +1


• Close to 0 indicate weak relationship
Structural • Significance of coefficients depends upon
standard error
Model • Two-tailed tests (t value should be greater
than the critical value)
• 10%(1.65); 5%(1.96); 1%(2.57)
• One-tailed tests (t value should be greater
than the critical value)
• 10%(1.28); 5%(1.65); 1%(2.33)
Coefficient of Determination
Values 0.25, 0.50, & 0.75
(R 2
)
0.2 is high in explaining
Values between 0 and 1 are weak, moderate, and
consumer behaviour
substantial

Evaluation
of
Structural Effect size f2
R2included – R2exluded / 1 - R2included

Model Values 0.02, 0.15, & 0.35 indicate small, medium, and large effects

(Contd…)
Blindfolding and Predictive Relevance Q2
Cross validated redundancy F2
Cross validated communality H2
approach (latent endogenous
approach (Manifest variables)
variables)
Mediation Analysis

• Mediator variable intervenes


between two variables
• A change in the exogenous
construct results in a change in the
mediator variable, which, in turn,
changes the endogenous construct
• Strong a priori
theoretical/conceptual support is
required to explore meaningful
mediating effects
• VAF = indirect effect / total effect
Mediation Analysis (Contd…)

Two types of non-mediation Three types of mediation


• Direct-only non-mediation (direct • Complementary mediation (indirect
effect is significant but not the and direct both are significant and
indirect effect) point in same direction)
• No-effect no mediation (neither • Competitive mediation (indirect and
direct nor indirect effect is significant direct both are significant and point
in opposite direction)
• Indirect-only mediation (indirect
effect is significant but not the
direct effect)
Practical Thank You
Demonstration

You might also like