Summary of Key Points

• SEM is not just a “statistical technique” but it integrates a number of different

multivariate techniques into one model fitting framework. SEM integrates the
various techniques such as measurement theory, factor (latent variable) analysis,
path analysis, regression and simultaneous equations.
• SEM confirms the causality among constructs and variables based on the evidence
of systematic covariation and helps the researcher to justify that relationships are
not false.
• SEM is a typical form of graphical modelling and therefore a system in which
relationship can be represented in either graphical or mathematical form.
• SEM is also known as covariance structure analysis, Analysis of moment
structures, analysis of linear structural relationships (LISREL), causal modelling.
• SEM helps to understand to what extent we are able to explain the structure of
relationships among the various exogenous and endogenous variables leading to
the final effect under investigation. Usually, we deal with linear structure but lots
of advancements are there to accommodate nonlinear structure also.
• SEM is considered as path analysis using latent variables.
• SEM is a framework for building and evaluating multivariate hypotheses about
multiple processes. It is not dependent on a particular estimation method.
• A model for SEM analysis must be derived from a sound underlying theory, and this
should include the development of measurement model specification and structural
model specification.
• SEM seeks to progress knowledge through cumulative learning. It advances the
knowledge from exploratory to confirmatory/hypothesis testing.
• SEM is considered to be a decision process which involves six steps (Hair et al.,
2012):
– Step 1: Definition of individual constructs (what items are to be used as measured
variables?).
– Step 2: Development of measurement model and path diagram.

© Springer Nature Singapore Pte Ltd. 2020 113

J. J. Thakkar, Structural Equation Modelling, Studies in Systems,
Decision and Control 285, https://doi.org/10.1007/978-981-15-3793-6
114 Summary of Key Points

– Step 3: Design of a study to produce empirical results (Check the adequacy

of sample size and select an appropriate estimation method and missing data
approach).
– Step 4: Assessment of the measurement model validity (Assess GOF and
construct validity of measurement model). If measurement model is found valid,
then proceed to stages 5 and 6 otherwise refine measures and design a new study.
– Step 5: Convert measurement model into a structural model.
– Step 6: Assessment of the structural model validity (Assess GOF and
significance, direction and size of structural parameter estimates). If
the structural model is valid, then draw the detailed conclusions and
recommendations otherwise refine model and test with new data.
• Underlying parameters in SEM
– Regression coefficients
– Variances and covariances.
• A hypothesized model is used to estimate these parameters for the population
(assuming the model is true)
– The parameter estimates are used to create a hypothesized variance/covariance
matrix
– The sample VC matrix is compared to the estimated VC.
• Combination of factor analysis and regression
– Continuous and discrete predictors and outcomes
– Relationships among measured or latent variables.
• Direct link between Path Diagrams and equations and fit statistics.
• SEM models include both measurement and path models.
• Principal component analysis versus factor analysis:
– In principal components analysis, the goal is to explain as much of the total
variance in the variables as possible. The goal in factor analysis is to explain
the covariances or correlations between the variables.
– Use principal components analysis to reduce the data into a smaller number
of components. Use factor analysis to understand what constructs underlie the
data.
– In principal components analysis, the components are calculated as linear
combinations of the original variables. In factor analysis, the original variables
are defined as linear combinations of the factors.
• Factor loadings represent how much a factor explains a variable in factor analysis.
• A communality is the extent to which an item correlates with all other items.
Higher communalities are better. If communalities for a particular variable are
low (between 0.0 and 0.4), then that variable may struggle to load significantly on
any factor.
Summary of Key Points 115

• Relationship between factor model and degree of freedom (DOF): If DOF =

0 (unique solution); DOF > 0 (multiple solutions); DOF < 0 (factor model is
indetermined).
• In SEM analysis, first develop measurement model (which is basically
confirmatory factor analysis) and then based on that develop structural model.
If structural model is valid, then only draw detailed conclusions.
• Multivariate regression versus path model: There is no interrelationship among
independent variables in multivariate regression model and also among dependent
variables. There is only a relationship between dependent and independent
variables. In path model, there is interrelationship among (within) dependent
variables and among (within) the independent variables.
• Endogenous (dependent latent) versus exogenous (independent latent)
variable: Exogenous variables are the one which are independent latent variables
and not affected by any other variable in the model (usually presented on the left-
hand side of the model) whereas others are endogenous variables which are either
affected by exogenous variables or endogenous variables.
• Goodness of fit indices: Measurement model and structural model both should
fit to the data—must be checked using goodness of fit indices like—chi-square,
root mean square residual, goodness of fit Index (GFI), normed fit index (NFI),
comparative fit index (CFI), incremental fit index (IFI).
• Absolute fit versus relative fit indices: Relative fit indices explain the goodness
of fit of the present model with respect to many other models possible whereas
absolute fit indices explain the goodness of fit for the model itself (under
consideration).
• Three most important issues for measurement model and structural model are:
(i) model identification; (ii) parameter estimation; (iii) model adequacy test.
• For the identification of measurement model, two conditions should be satisfied:
(i) order condition and (ii) rank condition. Order condition is bare minimum to be
satisfied and it satisfies the necessity and rank condition satisfies the sufficiency.
• Order condition: Number of parameters to be identified must be less than number of
non-redundant elements in the covariance matrix. Preferably model should be over
identified. Once model identification is done, you are ready for model parameter
estimation.
• Important assumption in measurement model or confirmatory factor analysis is
that the covariance between manifest variable or latent variable and error term is
zero.
• The output of measurement model is relationship between exogenous and
endogenous variables. This is utilized as an input to structural model.
• Structural model is basically path analysis. We conduct path analysis between
exogenous and endogenous variables whose relationships are determined based
on measurement model. We are not focusing on manifest variables.
• Every endogenous variable (right side of the structural model) may have direct
effect or indirect effect. We have to identify the total effect (direct and indirect) on
116 Summary of Key Points

the issue/problem/phenomena under investigation. This helps to rank or prioritize

the variables.
• The widely used software for SEM are: LISREL, MPlus, EQS, Amos, Calis, Mx,
SEPATH, Tetrad, R, Stata.
YouTube Links

S. No. Topic YouTube link Time (in min)

1 Structural equation modelling: https://www.youtube.com/ 25.32
what is it and what can we use it watch?v=eKkESdyMG9w
for? by Professor Patric Sturgis,
Department of Social Statistics
and Demography at the
University of Southampton,
published by National Centre for
Research Methods (NCRM)
2 Introduction to structural https://www.youtube.com/ 55.45
equation modelling watch?v=Hukuz1uNdjs
by Professor J. Maiti, IIT
Kharagpur published by National
Programme for Technology
Enhanced Learning (NPTEL)
3 Key ideas, terms and concepts in https://www.youtube.com/ 41.20
SEM watch?v=NOWdrfQVWAI
by Professor Patric Sturgis,
Department of Social Statistics
and Demography at the
University of Southampton,
published by National Centre for
Research Methods (NCRM)
4 Why use a structural equation https://www.youtube.com/ 11.28
model? watch?v= -m4ag3WQcCw
by Mr. Dan, Curran-Bauer
Analytics
5 Structural equation modelling https://www.youtube.com/ 12.47
using AMOS watch?v=fwrp0D7T1UI
by G N Satish Kumar, My easy
statistics

© Springer Nature Singapore Pte Ltd. 2020 117

J. J. Thakkar, Structural Equation Modelling, Studies in Systems,
Decision and Control 285, https://doi.org/10.1007/978-981-15-3793-6
Glossary of Key SEM Terminologies

Absolute fit indices It intends to examine overall goodness of fit for both the
structural and measurement models jointly. It does not compare the assumed
model with a null model (incremental fit measure) or attempts to adjust the
number of parameters in the estimated model (parsimonious fit measure). It is
a direct measure of how well the model specified by the researcher produces
the observed data. The SEM software packages compute multiple absolute fit
indices such as: chi-square statistic, goodness of fit (GFI), root means square
residual (RMSR), standardized root mean residual (SRMR), root mean square
error of approximation (RMSEA).
Coefficient of determination It is the fraction of variation explained by an equation
of a model.
Composite reliability The proportion of the variance of a variable which is not due
to measurement error.
Confirmatory factor analysis It is a factor analysis which finds the loading factors
to describe the correlation between the variables which are defined a priori. The
factors restricted.
Construct It is a broad concept which is defined conceptually with a theoretical
meaning. Constructs may be abstract and need not to be directly observable.
Examples: Intelligence or customer satisfaction.
Constructs are considered as unobservable which are represented by set of
variables. A typical mathematical relationship among variables represents a
construct.
Exogenous constructs are latent and equivalent to independent variables. These
are constructs that are determined outside of the model.
Exogenous constructs are latent equivalent to dependent variables.
Direct effect The effect of a variable on another due to the direct relationship
between them in the SEM.
Endogenous variable A variable is called endogenous if all the arrows point into
it and also called dependent variable. It is represented by a variate of dependent
variables.
Error Variance leftover after prediction of a measured variable.
© Springer Nature Singapore Pte Ltd. 2020 119
J. J. Thakkar, Structural Equation Modelling, Studies in Systems,
Decision and Control 285, https://doi.org/10.1007/978-981-15-3793-6
120 Glossary of Key SEM Terminologies

Exogenous variable A variable is called exogenous if the paths originate from it and
also called independent variable. These are the constructs which are determined
by factors outside of the model.
Exploratory factor analysis It is a factor analysis which finds the loading factors
to describe the correlation between the variables. In this type, the factors are not
restricted and all are connected to each other.
Goodness of fit index (GFI) It measures how well a specified model reproduces
the covariance matrix among the observed variables. This means to what extent
the similarity exists between the observed and estimated covariance matrices.
Imputation It tries to estimate the values of missing data based on the valid values
of the other variables. This is accomplished by identifying relationships with the
valid sample and using this identified relationship for estimating the value of a
missing observation.
Incremental fit indices These are goodness of fit indices used to examine how
well a specified model fits relative to an another baseline model. The baseline
model is considered as the null model which assumes that all the measured
variables are unrelated to each other. It supports the other goodness of fit
indices such as absolute fit and parsimonious fit. The SEM programmes compute
multiple incremental fit indices such as normed fit index (NFI), comparative fit
index(CFI), Tucker Lewis index(TLI), relative noncentrality index (RNI).
Indirect effect The effect of a variable on another due to an intermediate variable
between them in the SEM.
Instrumental variable A variable which introduces exogenous variability into an
endogenous variability.
Latent variable A variable in the model which cannot be measured. It can also be
called factor or unmeasured variable.
Loading factors It is a factor which indicates the effect of one variable on another.
Measurement model It indicates how measurement variables together represent
constructs.
Measurement model The part of the model that relates indicators to latent factors.
It is the factor analytic part of SEM.
Model fit The ability of an overidentified model to reproduce the correlation or
covariance matrix of the variables.
Model identification Identification of the model is to know whether the solution
exists for a model or not. A model is said to be identified if there exists a unique
solution for all parameters.
Modification index It is a score test for adding paths where none occur.
Multicollinearity It explains to what extent a particular construct can be explained
by the other constructs in the analysis. An increase in multicollinearity raises the
complexity in interpretation as a particular construct has many interrelationships
with other variables and it is difficult to determine the effect of this construct in
the analysis.
Observed variable A variable in the model which is present in the data set and
measured. It is also called as an indicating variable for a latent variable.
Glossary of Key SEM Terminologies 121

Over fit model It is a model which is very much exactly related to a particular set of
data and fails to fit additional data of the system. It has a limitation in predicting
the future reliably. It includes more parameters (typically known as noise factors
or residual variations) than the parameters necessary to fit the data and represent
an underlying model structure.
Path diagram A graphical representation of the structural equation model using
ellipses, circles, rectangles single-headed and double-headed arrows.
Parsimonious fit indices It indicates the overall goodness of fit by establishing
relationship between degree of model fit and estimated coefficient. It enables
a researcher to correct the model for overfitting by evaluating the parsimony
ratio of the model. It complements the other two goodness of fit indices such as
absolute fit and incremental fit.
Parsimonious ratio It compares the degree of freedom of specified model and the
total number of degrees of freedom available. It indicates to what extent the
model utilizes the total number of degrees of freedom available and helps in the
examination of overfitting the model with additional relationships that ensure
only marginal gains in model fit.
Path analysis It determines the strength of the paths indicated in path diagram using
simple bivariate correlations and estimates the relationships in SEM model.
Path diagram It is a graphical and visual representation of all the relationships
among the model’s constructs. A straight arrow is used to indicate the dependence
relationships with an arrow emanating from the predictor variable and the arrow
head pointing to the dependent construct or variable. Usually, curved arrows
are used to represent correlations between constructs or indicators without any
causation.
Path model This is the part of the model that relates variables or factors to one
another (prediction). If no factors are in the model then only path model exists
between indicators.
Recursivity It is a state where all causal is unidirectional and disturbances are
uncorrelated.
Reliability It is used to indicate the internal consistency of a set of indicators of
a latent construct on the basis that how highly indicators are interrelated. It
indicates an ability of the indicators to measure the same thing (latent construct).
The deviation in measurement indicates the degree of error.
Residual It represents the difference between the actual and estimated value for
any relationship. Typically, in SEM analysis, it is the difference between the
observed and estimated fitted covariance matrices.
Root means square residual (RMSR) It is the error in the prediction of each
covariance term which indicates the residual created. RMSR is the square root
of the mean of these squared residuals which represents an average of the
residuals between individual observed and estimated covariance and variance
terms. A standardized value of RMSR is known as standardized root mean
residual (SRMR). Lower values of SRMR and RMSR indicate the better fit
and higher value represents the poor fit.
122 Glossary of Key SEM Terminologies

Structural model The model in which the parameters are a description of the
scenario and also representing causal effects between the variables. It indicates
how the constructs are associated with each other. It is also known as theoretical
model or causal model.
Structural equation modelling (SEM) It is a multivariate analysis technique
which integrates two methodologies such as factor analysis and multiple
regression analysis to simultaneously examine series of interrelated dependence
relationships among the measured variables and latent constructs (variables) as
well as between several latent constructs.
Under fit model It fails to adequately capture the underlying structure of the data
which has some parameters missing that are necessary to ensure the correctness
of the model. Typically, this occurs while fitting a linear model to nonlinear data.
Variable Actual items that are measured using a survey, observation or some other
measurement device. It is created by developing constructs into measurable form.
It represents a characteristic which varies and has at least two possible values.
Variables are considered as observable items in the sense that we can get a direct
measurement of them.
Example: Productivity, age of the students, weight, etc.
Variance and covariance matrix The matric which represents the variance and
covariance between two variables. The diagonal elements represent the variance
and the other elements represent the covariance between the variables in
respective row and column of that element.
Bibliography and Further Reading

Anderson, T. W., & Rubin, H. (1956). Statistical inference in factor analysis. In J. Neyman (Ed.),
Proceedings of the third Berkeley symposium on mathematical statistics and probability (Vol. V,
pp. 111–150). Berkeley: University of California Press.
Arhonditsis, G. B., Paerl, H. W., Valdes-Weaver, L. M., Stow, C. A., Steinberg, L. J., & Reckhow, K.
H. (2007). Application of Bayesian structural equation modeling for examining phytoplankton
dynamics in the Neuse River Estuary (North Carolina, USA). Estuarine, Coastal and Shelf
Science, 72(1–2), 63–80.
Azam, A., Qiang, F., Abbas, S. A., & Ibrahim Abdullah, M. (2013). Structural equation modeling
(SEM) based trust analysis of Muslim consumers in the collective religion affiliation model in
e-commerce. Journal of Islamic Marketing, 4(2), 134–149.
Cudeck, R., Du Toit, S., & Sörbom, D. (Eds.). (2001). Structural equation modeling: Present and
future. A Festschrift in honor of Karl Jöreskog. Lincolnwood, IL: Scientific Software International.
de Beer, L. T., Rothmann Jr., S., & Pienaar, J. (2016). Job insecurity, career opportunities,
discrimination and turnover intention in post-apartheid South Africa: Examples of informative
hypothesis testing. The International Journal of Human Resource Management, 27(4), 427–439.
Froese, F. J., & Xiao, S. (2012). Work values, job satisfaction and organizational commitment in
China. The International Journal of Human Resource Management, 23(10), 2144–2162.
Gopal, P. R. C., & Thakkar, J. J. (2016). Sustainable supply chain practices: An empirical
investigation on Indian automobile industry. Production Planning and Control, 27(1), 49–64.
Hair, J. F., Black, W. C., Babin, B. J., Anderson, R. E., & Tatham, R. L. (2012). Multivariate data
analysis (6th ed.). London: Pearson.
Hair, J. F, Jr., Sarstedt, M., Hopkins, L., & Kuppelwieser, V. G. (2014). Partial least squares structural
equation modeling (PLS-SEM). European Business Review, 26(2), 106–121.
Hershberger, S. L. (2003). The growth of structural equation modeling: 1994–2001. Structural
Equation Modeling, 10(1), 35–46.
Jain, N. K., Kamboj, S., Kumar, V., & Rahman, Z. (2018). Examining consumer-brand relationships
on social media platforms. Marketing Intelligence & Planning, 36(1), 63–78.
Jin, N., Lee, S., & Gopalan, R. (2012). How do individual personality traits (D) influence perceived
satisfaction with service for college students (C) in a casual restaurant setting (I)?: The CID
framework. Journal of Hospitality Marketing & Management, 21(6), 591–616.
Kharouf, H., Lund, D. J., & Sekhon, H. (2014). Building trust by signaling trustworthiness in service
retail. Journal of Services Marketing, 28(5), 361–373.
Khoreva, V., & Tenhiälä, A. (2016). Gender differences in reactions to injustice. Journal of
Managerial Psychology, 31(4), 790–804.
Lee, H., & Kim, J. (2018). A structural equation model on Korean Adolescents’ excessive use of
smartphones. Asian Nursing Research.

© Springer Nature Singapore Pte Ltd. 2020 123

J. J. Thakkar, Structural Equation Modelling, Studies in Systems,
Decision and Control 285, https://doi.org/10.1007/978-981-15-3793-6
124 Bibliography and Further Reading

Li, C., & Jiang, S. (2018). Social exclusion, sense of school belonging and mental health of migrant
children in China: A structural equation modeling analysis. Children and Youth Services Review,
89, 6–12.
Lu, C. S., Lai, K. H., & Cheng, T. E. (2007). Application of structural equation modeling to
evaluate the intention of shippers to use Internet services in liner shipping. European Journal of
Operational Research, 180(2), 845–867.
Mishra, A., Maheswarappa, S. S., & Colby, C. L. (2018). Technology readiness of teenagers: A
consumer socialization perspective. Journal of Services Marketing.
Panda, S., & Rath, S. K. (2017). The effect of human IT capability on organizational agility: An
empirical analysis. Management Research Review, 40(7), 800–820.
Pokharel, M. P., & Ok Choi, S. (2015). Exploring the relationships between the learning organization
and organizational performance. Management Research Review, 38(2), 126–148.
Prasertratana, S., Sanratana, W., & Somprach, K. O. (2014). Distributed leadership: Structural
equation model. Procedia-Social and Behavioral Sciences, 116, 1610–1614.
Rezaei, M., & Ghazanfari, F. (2016). The role of childhood trauma, early maladaptive schemas,
emotional schemas and experimental avoidance on depression: A structural equation modeling.
Psychiatry Research, 246, 407–414.
Shih, W. L., & Tsai, C. Y. (2016). The effects of knowledge management capabilities on perceived
school effectiveness in career and technical education. Journal of Knowledge Management, 20(6),
1373–1392.
Sin, A. B., Zailani, S., Iranmanesh, M., & Ramayah, T. (2015). Structural equation modelling on
knowledge creation in Six Sigma DMAIC project and its impact on organizational performance.
International Journal of Production Economics, 168, 105–117.
Singh, A. K. (2015). Modeling passengers’ future behavioral intentions in airline industry using
SEM. Journal of Advances in Management Research, 12(2), 107–127.
Tan, L. P., & Wong, K. Y. (2015). Linkage between knowledge management and manufacturing
performance: A structural equation modeling approach. Journal of Knowledge Management,
19(4), 814–835.
Tummers, L., Kruyen, P. M., Vijverberg, D. M., & Voesenek, T. J. (2015). Connecting HRM
and change management: The importance of proactivity and vitality. Journal of Organizational
Change Management, 28(4), 627–640.
Van Poucke, E., Matthyssens, P., & Weeren, A. (2016). Enhancing cost savings through early
involvement of purchasing professionals in sourcing projects: Bayesian estimation of a structural
equation model. Journal of Purchasing and Supply Management, 22(4), 299–310.
Vázquez-Bustelo, D., Avella, L., Fernández, E. (2007). Agility drivers, enablers and outcomes:
Empirical test of an integrated agile manufacturing model. International Journal of Operations
& Production Management, 27(12), 1303–1332.
Wang, W., Wang, Y., Xiao, C., Yao, X., Yang, Y., Yan, H., & Li, S. (2018). Psychological pathway
to suicidal ideation among people living with HIV/AIDS in China: A structural equation model.
Psychiatry Research, 260, 255–261.