A Review of PLS-SEM as Statistical Approach for
Business Research
Dr Sarvesh Kumar
Assistant Professor,
School of Commerce and Management
Studies,
Central University of Himachal Pradesh,
Dhauladhar campus— II, Dharamshala,
Himachal Pradesh
E-mail- sarvesh_hcu@yahoo.co.in
Vikas Kumar Tyagi
Assistant Professor,
Department of Management Studies,
Panipat Institute of Engineering and
Technology,
Samalkha, Haryana
E-mail- vktyagi@outlook.in
Dr. Yoginder Singh Kataria
Professor,
Faculty of Commerce and Management
SGT University, Gurugram, Haryana
E-mail- yoginderkataria@gmail.com
Abstract:
Structural Equation Modeling (SEM), a
Second generation multivariate data analysis technique isin use for data
analysis, especially hypothesis testing from a long time. There are different
approaches to SEM used in research. The objective
of this research paper was to review different Structural Equation Modelling
techniques used in Business Research, with special reference to variance based
Partial Least Square Structural Equation Modeling (PLS-SEM). PLS-SEM is more
sensitive, simpler, and powerful statistical technique for data analysis. It is
mainly used in theory development for exploratory purposes and has less strict
assumptions. Severaltypes of research have been done in the past on this
subject; this paper compiles all the ideas by reviewing them and coveringthem
in the least complicated form. It would especially help the researchers and
practitioners in understanding the differences between different approaches to
SEM and its applications.
KEYWORDS: Structural Equation Modeling,
Partial Least Square, Covariance, Variance.
JEL CLASSIFICATION:
C14, C31, C88
Introduction:
Volatility, Uncertainty, Complexity, and
Ambiguity(VUCA) originated in the US Military (Whiteman, 1998) as cited by Bennett & Lemoine (2014). It has
become a synonym of constant change in dynamic business research scenario. Embedding
mathematical analysis simplifies decision making in a dynamic business
environment. PLS-SEM has evolved as a handy tool for researchers. This tool
also helps researchers and decision makers to reach confident decisions concerning
their defined problems. Sewall Wright first developed Path analysis models in
the year 1921 (Wolfle, 1980). Structural Equation Modeling (SEM) is a second generation multivariatedata analysis technique (Elangovan & Rajendran,
2015). SEM has advantage of analyzing multiple layers of links between
independent (IV) and dependent variables (DV) simultaneously over first generation
regression models like Linear Regression, Analysis of variance (ANOVA), and Multivariate analysis of
variance (MANOVA); it can. Researchers use SEM extensively for hypothesis
testing (Bagozzi & Yi, 1988).
There are several, but the researchers
most commonly use two approaches to SEM. The first approach is the more widely
used and is older than PLS-SEM; it is Covariance-based SEM (CB-SEM), which is
more of confirmatory and conclusive. Covariance is the extent of how much two
variables change together or how well the variables are jointly related (Davis
& Pecar, 2013). The second approach, Variance based Partial Least Squares
SEM (PLS-SEM) is more exploratory in nature and was less commonly used till recently.
Variance is the measure of the dispersion of the observations, to check how
dispersed the data values are about the mean values (Davis & Pecar, 2013). Hwang&
Takane, (2004) came up with the third approach, i.e. Generalized Structured
Component Analysis (GSCA), after these two methods. The researcherhas kept GSCA outside the purview of this
research.
In the quest
of statistical significance, during different time horizons, different types of
methods to interpret data have evolved, solving the problem of that time. Over
the years, the inherent gaps of existing methodology paved the path of newer
methodology whose structure has been built by refining current methods. Considering
one technique to be better than the other would not be right, both have their
advantages and disadvantages, discussed later in this paper.
There are several studies done on this
subject, giving their opinions and different aspects of the subject; this paper compiles all the ideas by reviewing them. This research paper would
help in understanding the differences between the concepts. This paperis an
essential guide reducing complexity, for learning it in more details researcher
suggests to go through researches such as Gaskin (2018); Hair Jr, Hult, Ringle,
& Sarstedt (2016) and Byrne, (2016).
In VUCA times PLS-SEM is gaining lot of prominence
in business research as popular choice of business research methods.
Internationally much work has been done by using the concept PLS-SEM, such as
studies done in different fields such as by Aryanto, Fontana, & Afiff
(2015); Astrachan, Patel, & Wanzenried (2014); Hair, Sarstedt, Pieper,
& Ringle (2012); Okazaki, Mueller, &Taylor (2010); and Schwaiger,
Sarstedt, & Taylor (2010) in Marketing Research field. Peng & Lai (2012)
in the area of Operations Management. Chen, Preston, & Xia (2013) and
Hoffmann, Schiele, & Krabbendam (2013) in the field of Supply Chain
Management. Kallunki, Laitinen, & Silvola (2011) in Accounting Research. In
India- Atulkar & Kesari (2018); Kesari, B., & Atulkar, S. (2016);
Kamath, Rodrigues, & Desai (2016); Shanmugapriya, & Subramanian (2015);
Venkatesh, Sykes, & Venkatraman (2014); Seetharaman, Bajaj, Raj, &
Saravanan (2013); did studies by using PLS-SEM in the area of business
research.
In this paper, the introduction is
followed by the comparison between both PLS-SEM and CB-SEM and followed by the
advantages of one over another and situations when each one of them should be
selected. In the end, the conclusory remarks about the issue are mentioned.
Comparative Analysis between PLS-SEM and CB-SEM
Karl Joreskog developed CB-SEM (Joreskog,
1970) in behavioural sciences, whereas PLS-SEM was developed initially by Herman
Wold (1974) in social sciences. Sosik, Kahai, & Piovoso, (2009) considered
CB-SEM as hard modelling for theory
testing whereas considered PLS-SEM as
soft modelling approach for the theory
development. Tenenhaus (2008) called the
PLS-SEM as component based SEM, just as the other one is called Covariance-based approach.
The most significant difference between
them lies in their respective objectives for which they are used. CB-SEM helps
inevolvingtheoretical covariance matrix by estimating model parameters with an
objective of minimizing the differences between theoretical covariance matrix
and the estimated covariance matrix, without focusing on the explained
variance. Whereas, PLS-SEM is used with an objective ofmaximizing the explained
variance of the dependent latent
construct in business research (Hair, Ringle, & Sarstedt, 2011). This is
the reason why majority of researchers do
not accept standard goodness-of-fit statistic of PLS-SEM.
CB-SEM is the more popular technique as
it provides flexibility of using various software’s like Mplus, EQS, LISREL, andAMOS.
Whereas, PLS-SEM is a less popular
technique despite being a more robust
estimator (Reinartz, Haenlein, & Henseler, 2009). The main reason for its lower popularity is also because the software on which it runs like: SmartPLS by
Ringle, Wende, and Will in 2005 and PLS-Graph
by Chin in 2003 were developed much later. Other PLS Software’s such as
VisualPLS, WarpPLS, and ‘R’ statistical software package, can also be used to
run PLS-SEM. Although VisualPLS and
PLS-Graph have graphical interfaces they have not received any significant
updates since their release, and ‘R’ requires a bit advanced programming
language skills, making SmartPLS most used and popular software for PLS
applications. Its 2.0 version is available freely, making it the best-suited software for the purpose (Wong,
2013).
Table
1: Studies comparing CB-SEM vs PLS-SEM, performed in different fields:
|
Author(s) |
Area
of Research |
Conclusions |
|
Richter,
Sinkovics, Ringle, & Schlaegel (2016) |
International
Business Research |
Out of 424 studies reviewed which used
SEM, 379 were using CB –SEM, and the rest 45 were using PLS-SEM. They were
used due to their lower sample sizes and data measurement issues in place of
their objectives. Studies still don’t reap the benefits
of PLS-SEM to its full Extent. |
|
Kaufmann
& Gaeckler (2015) |
Supply
Chain Management |
Use of PLS-SEM has magnified in the
SCM recently, but most of them didn’t follow the standards of the technique. They found CB-SEM to be better for the
subject if its assumptions are met. |
|
Astrachan,
Patel, & Wanzenried (2014) |
Family
Firms Research |
Found PLS-SEM better than CB-SEM for
the studies in their area of research. PLS-SEM enables the extension of more
indicator variables, whereas, CB-SEM explains variance better, but in the
case of Non-normal data, CB-SEM gives inflated R2. |
|
Hair,
Sarstedt, Ringle, & Mena (2012) |
Marketing
Research |
PLS-SEM has become more widely used in
Marketing Research. However, this has been misunderstood as it lacks in the
standard textbooks. It is a robust technique but should not be applied to
ditch the assumptions of CB-SEM. |
|
Peng
& Lai (2012) |
Operation
Management Research |
PLS-SEM is most widely used in the
field of the information system, and not as widely in the area of Operation
Management field, where still CB-SEM is used. They found CB-SEM to be superior, but
only if its assumptions are met otherwise, PLS-SEM should be opted. |
|
Gefen,
Rigdon, & Straub (2011) |
Administrative
and Social Science Research |
Found that it would not be right to
mention one of them to be better than another. However, they should be opted
as per their objectives. PLS-SEM should be used in exploratory studies,
whereas CB-SEM in confirmatory studies. |
|
Lee,
Petter, Fayard, & Robinson (2011) |
Accounting
Research |
PLS-SEM is commonly used in Social
Sciences but not in Accounting discipline where traditional regression is
used. Studies in this field are yet to avail its benefits to its full extent.
|
|
Fornell
& Bookstein (1982) |
Marketing
Applications |
Found PLS-SEM to be a more feasible
option. Although CB-SEM was found to give statistically precise results but
need to follow strict assumptions and require a larger sample size for
accurate results. Whereas, PLS-SEM have prediction accuracy even with smaller
sample sizes and can also deal with larger models with many variables. |
Source: Literature Review
Advantages of one over another
According to Hair, Ringle, & Sarsted
(2011) PLS-SEM has a higher level of
statistical power as compared to CB-SEM; generates better path coefficients and
significance level and is more sensitive in detecting relationships (Sosik,
Kahai, & Piovoso, 2009); is simpler in nature (Tenenhaus, 2008), yet can
deal with high model complexity (Hair, Ringle, & Sarstedt, 2011). PLS
techniques can work even when there are just one or two items per construct,
unlike CB-SEM. It can deal with both formative and reflective measurements
models more easily than CB-SEM can, which requires relatively more complex
rules (Hair, Ringle, & Sarstedt, 2011 and Sosik, Kahai, & Piovoso,
2009).
PLS methods are non-parametric
techniques (Nagarajan, Savitskie, Ranganathan, Sen, & Alexandrov, 2013), unlikeCo-variance based methods, which are parametric.
Therefore, unlike CB Techniques, researchers do not have to satisfy any sets of
assumptions before the application of PLS techniques. Model specifications or
data in PLS-SEM do not use any limiting assumptions. While,multivariate
normality of data can work well on non-normal data; PLS algorithm adjusts a
non-normal data according to the central limit theory (Cassel, Hackl, &
Westlund, 1999); minimum sample size as it can work with a small and much wider range of sample sizes
(Diamantopoulos & Siguaw, 2013 and Ringle, Sarstedt, & Straub, 2012).
Observational independence and interval scaled data as PLS-SEM can be applied
even when the data is non-independent or
is in Ordinal or Nominal scale (Sosik, Kahai, & Piovoso, 2009). According
to Wong (2013), it is challenging to find the data that meet all these
assumptions. Moreover, as commented by
Wold (1982) as cited in Hair, Ringle, & Sarstedt (2011), the informational
and distributional requirements or assumptions for CB-SEM are unrealistic. They
are making PLS Techniques more realistic.
PLS-SEM is more suitable on the smaller
sample sizes compared to CB-SEM (Wong, 2013); (Sosik, Kahai, & Piovoso,
(2009); (Tenenhaus, 2008) and (Marcoulides & Saunders, 2006). Ringle, Sarstedt, & Straub (2012) reviewed/meta-analyzed
204 of studies which were done using PLS-SEM, they found that the average
sample size in those studies was 238.12, and the median was 198. In addition to
that, it was evident through the study
that 33.8 studies used this method only because of having low sample sizes.
Tenenhaus, Pages, Ambroisine, & Guinot (2005) even did research based on just six subjects.
Even though PLS-SEM has the edge over
CB-SEM, CB-SEM also has few comparative advantages over PLS-SEM. PLS-SEM allows
for only recursive relationships in the structural models with no causal loops
allowed. Whereas, CB-SEM does not have any such restriction and can also work
well with the non-recursive models; CB-SEM can deal with bidirectional
relationships unlike in PLS-SEM where relationships tend to unidirectional only
(Hair, Ringle, & Sarstedt, 2011). CB-SEM gives better global goodness of
fit criterion (Hair, Ringle, & Sarstedt, 2011) compared to PLS-SEM with no
global measure of goodness of model fit.
As PLS-SEM output does not give any
overall model fit in, therefore Goodness of fit (GoF) proposed by Tenenhaus,
Vinzi, Chatelin, & Lauro (2005) can be used to assess the structural model.
The geometric mean (G.M.) of the Average Variance Extracted (AVE) and the
average Coefficient of Determination (R2) is used for the
calculation of Goodness of fit value. Wetzels, Odekerken-Schröder, & Van
Oppen (2009) proposed the cut-off values for assessing the result of GoF
analysis as ‘GoF=0.10 (small); GoF=0.25
(medium); and GoF=0.36 (large)’. Still,
these cut off values are not widely accepted by the majority of researchers.
Selection between PLS-SEM Vs CB-SEM
As the researcher has already discussed
the positive and negative aspects of both the methods in this study; now will
discuss the situations in which they should be applied.
CB-SEM should be selected when the prior
theory is strong and additional testing &validation are the research objectives.
CB-SEM should also be applied when either the model is non-recursive, or when the assumptions of parametric tests are
fulfilled, or when bidirectional paths are used in the model (Wong, 2013; and
Hair, Ringle, & Sarstedt, 2011).
Whereas, PLS-SEM should be selected, if
the research is more of exploratory in nature or if it is an extension to an
existing theory. PLS-SEM should be used with the goal of predicting key target and
driver constructs, i.e. more for theory development rather than theory
confirmation (Hair, Ringle, & Sarstedt, 2011), notably when the sound
theory base is missing. PLS-SEM should be used when available theories on the
subject are insufficient (Wong, 2013). Especially during the early stage of
theory development (Sosik, Kahai, & Piovoso, (2009). PLS-SEM is strongly
prescribed when the model under study are complex or when formative constructs
are part of the structural model (Ringle, Sarstedt, & Straub, 2012). Moreover,
of course, its usage becomes must when
the underlying assumptions of CB-SEM are violated (Reinartz, Haenlein, &
Henseler, 2009), or when there is single item per construct (Ringle, Sarstedt,
& Straub, 2012).
However, in the condition when the sample size is considerably large, then both
the techniques give similar results,
irrespective of the assumptions or anything else (Sosik, Kahai, & Piovoso,
2009). As according to the Figure: 1, PLS-SEM and Figure: 2, CB-SEM same
manifest variables (USEF3, EOU3, BI3, ATT2, USE1) had maximum loadings for the
Latent Variables (USEF, EOU, BI, ATT, USE). Also, Path EOUàUSEF
followed by USEFàBI had maximum and second highest
coefficient values, and the sequence of other paths were also the same in both the models. The results were the same; only the values were different. The
reason for this could be because the
dataset was huge as in this hypothetical it was of 1,190 responses.
Figure:
1 Partial least squares (PLS) results
Source: (SmartPLS, 2018)
Figure:
2 CB-SEM Maximum likelihood (ML) based results (AMOS; standardized
coefficients)
Source: (SmartPLS, 2018)
Steps to be followed in PLS-SEM concerning
reflective scale:
First of all, the objectives should be
clearly defined and if they suit for
PLS-SEM, then only it should be applied. Additionally,
the assumption of the CB-SEM should be
checked before proceeding further with the PLS-SEM
if the data set fails to satisfy the assumptions than one would be forced to
move towards the PLS Techniques. One
should mention the reasons for using PLS Technique in their research.
There are two sub-models in a PLS
Structural Equation Model; (Sosik, Kahai, & Piovoso, 2009). The first component of PLS-SEM model is the
structural model or inner loop, having endogenous constructs which are
explained by relationships in structural model with the other constructs (Hair,
Ringle, & Sarstedt, 2011). It stipulates the relationships among the
independent and dependent latent variables. The second
component is an outer loop or measurement
model having exogenous constructs with no structural path relationship between
them (Hair, Ringle, & Sarstedt, 2011). It represents the relationships
between the latent variables and their manifest variables.
It cannot be said that there are no
restrictions on the sample size of the data
set while performing PLS Techniques. Sample size in case of PLS-SEM should be atleast
equal to or more than ten times highest formative variables measuring one
construct. (Hair, Ringle, & Sarstedt, 2011). Alternatively,minimum ten
times the highest statistical paths directed at a latent construct in the
structural model. According to Marcoulides & Saunders (2006) as cited in
Wong (2013); minimum sample size should be 91 ifmaximum
number of arrows pointing towards a latent variable is 10.
Checking validity and reliability is the mostcrucial step while performing PLS Techniques:
• For
Internal consistency/reliability (Nunnally
& Bernstein, 1994): in place of Cronbach alpha, Composite
Reliability is the better method in PLS-SEM, as it does not assume all the
indicators to be equally reliable (Bagozzi & Yi, 1988). Composite
reliability of more than 0.60 isregarded as
satisfactory.
• For
indicator reliability (Hair, Ringle, & Sarstedt, 2011 and Bagozzi & Yi,
1988): Indicator loadings should be more than 0.70 and loadings that are less
than 0.4 should not be included inreflective
scale.
• For
Convergent validity (Hair, Ringle, & Sarstedt, 2011) and (Bagozzi & Yi,
1988): Average Variance Extracted (AVE) should be more than 0.5,meaning that
the latent variable explains more than half of the variance of its indicators.
• For
Discriminant validity (Hair, Ringle, & Sarstedt, 2011): all cross-loadingsshould be lowerthan indicator loadings and AVE should be more than
the construct’s highest squared correlations with any other latent construct
(Fornell-Larcker Criterion) (Fornell & Larcker, 1981).
Measuring ‘Coefficient of determination’
(R2),i.e.model’s predictive
accuracy is primary evaluation criteria for the structural model. R2 values of 0.2 are
considered high in consumer behaviour
studies. Values of 0.7, for endogenous
latent construct is called as substantial, 0.5 as moderate and 0.25 asweek
(Hair, Ringle, & Sarstedt, 2011). Additionally, PLS Algorithm must converge
in maximum of 300 iterations, if it does not converge
in 300 iterations it would have meant that data were abnormal due to the
reasons such as the sample size could be too small, evidence of the existence of outliers, data having too many
identical values in indicator and this would require further investigation (Wong, 2013).
Bootstrapping, which is a non-parametric
method, allows for the testingnull hypothesis
that a coefficient equals to zero, through this,
the significance of the coefficient can
be analyzed. Bootstrapping should be
done with samples of at least 5,000 wherenumber
of cases should not be less than the number
of original observations. Omission distance (d) should be chosen between 5 and
10. Critical t-values for a two-tail test
are 1.65 (significance at the level of 10 percent),
1.96 (significance at the level of 5 percent),
2.58 (significance at the level of 1 percent)
(Hair, Ringle, & Sarstedt, 2011).
Model’s capacity to predict is significant
(Rigdon, 2014) and can be
measured through Stone-Geisser’s (Q2) (Stone, 1974) obtained through
Blindfolding, which gives ‘cross-validated
redundancy’ and ‘cross-validatedcommunality’.
Hair, Ringle, & Sarstedt (2011) suggested using
cross-validated redundancy (Q2) of endogenous latent variable
and its value to be more than zero for explaining the construct’s predictive relevance.
With the significance,
it is also f2 value indicates the effect of the construct removed
for a particular endogenous construct. The values of 0.02 represent small, 0.15
medium and 0.35 large effects (Cohen, 1988). If an exogenous construct strongly
contributes to explaining an endogenous construct, the difference between R2
included and R2 excluded should be high, leading to high effect size (f2 value).essential
to measure the magnitude of the influence, which can be done through Cohen’s f2 or Effect size.
f2 = (R2AB –R2A)
/ (1- R2AB)
Where;
|
R2A |
= |
variance accounted for in the
population by variable set A |
|
R2AB |
= |
Variance accounted for in the
population by variable set A and B Together |
Conclusion:
Data itself is not the solution to any
problem; it is the analysis of data which shows the decision path to any
manager. For the analysis, several techniques could be used. Comparisons among
the techniques would be wrong as none of them is better than another. Their
usage should depend upon the objectives of any particular study. In a
situation, one could be better whereas in another case the other one. The
relation between both is more complementary rather than competitive. As also
found by Tenenhaus (2008) and cited in
Hair, Ringle, &
Sarstedt(2011) –When the study is using suitable measures and data or
when CB-SEM assumptions were met then in that situation both the approaches
practically yield the same results. That is if before the application of
PLS-SEM if measurement model characteristics are checked,then it will give similar results as of CB-SEM
(Hair, Ringle, & Sarstedt, 2011).
If precisely discussing the strength of
PLS, then it is statistically more robust, sensitive, and simpler with less
strict assumptions. However, its inability to deal with non-recursive
relationships and causal loops; and the absence of global goodness of fit acts
as its weakness.
In the dynamic business environment,
VUCA is quite high, and almost every data in the social sciences follows
Pareto’s principle. It does not showcase normality and is mainly skewed, also
as the life cycle of every product and service is shortening due to
technological advancements, it becomes imperative to finish the marketing
research very rapidly. Therefore, it becomes difficult to fulfil all the
assumptions of the CB-SEM and to have a large sample size, which acts as an
opportunity for the usage of PLS. seeing the dynamic nature of the ever-changing
business scenario, this methodology facilitates the researcher with appropriate
judgmental decision choices.
However, every researcher should first
identify her or his objectives and then decide the better-suited method for her or his study. Dijkstra &Henseler (2015)
gave consistent PLS (PLSc) model, which is an extension of PLS and is comparable
to covariance-based SEM, in future,the researchers suggest a study
differentiating PLSc and CB-SEM. Researchers believe that PLS-SEM usage has
been overly used or indeed misused due to its simplicity or they believe that
it will require less pain due to lesser strict assumptions. Whereas, they must
be considering their objectives as their selection criterion to choose any
method.
References:
Aibinu, A. A.,
& Al-Lawati, A. M. (2010). Using PLS-SEM technique to model construction
organisations' willingness to participate in e-bidding. Automation in
construction, 19(6),
714-724.
Annamalai, C.,
&Ramayah, T., (2011). A review of ERP implementation in India. International Journal of Business and
Systems Research, 5(4), 406-421.
Aryanto, R.,
Fontana, A., &Afiff, A. Z. (2015). Strategic human resource management,
innovation capability and performance: An empirical study in Indonesia software
industry. Procedia-Social and Behavioral Sciences, 211, 874-879.
Astrachan, C.
B., Patel, V. K., &Wanzenried, G. (2014). A comparative study of CB-SEM and
PLS-SEM for theory development in family firm research. Journal of Family
Business Strategy, 5(1),
116-128.
Atulkar, S.,
&Kesari, B. (2018). Impulse buying: A consumer trait prospective in context
of central India. Global Business Review, 19(2), 477-493.
Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural
equation models. Journal of the Academy of Marketing Science, 16(1), 74-94.
Baumgartner, H., & Homburg, C. (1996). Applications of structural equation
modeling in marketing and consumer research: A review. International
journal of Research in Marketing, 13(2), 139-161.
Bennett, N., &Lemoine, G. J. (2014). What a difference a word makes:
Understanding threats to performance in a VUCA world. Business Horizons, 57(3),
311-317.
Blunch, N.
(2008). Introduction to structural equation modelling using SPSS and
AMOS. Sage.
Cassel, C.,
Hackl, P., &Westlund, A. H. (1999). Robustness of partial least-squares
method for estimating latent variable quality structures. Journal of
Applied Statistics, 26(4),
435-446.
Caniëls, M. C.,
Gehrsitz, M. H., &Semeijn, J. (2013). Participation of suppliers in
greening supply chains: An empirical analysis of German automotive
suppliers. Journal of Purchasing and supply management, 19(3),
134-143.
Chen, D. Q.,
Preston, D. S., & Xia, W. (2013). Enhancing hospital supply chain
performance: A relational view and empirical test. Journal of
Operations Management, 31(6), 391-408.
Davis, G.,
&Pecar, B. (2013). Business Statistics using Excel. Noida: Oxford
University press.
Diamantopoulos,
A., &Siguaw, J. A. (2013). Introducing LISREL: A guide for the
Uninitiated. Sage.
Dijkstra, T. K.,
&Henseler, J. (2015). Consistent Partial Least Squares Path Modeling. MIS Quarterly, 39(2).
Elangovan, N.,
&Rajendran, R. (2015). Structural equation modeling-A second-generation
multivariate analysis. Indian Business
Management (pp. 33-54.). Tamil Nadu: Sri Ramakrishna Institute of
Technology.
F. Hair Jr, J.,
Sarstedt, M., Hopkins, L., & G. Kuppelwieser, V. (2014). Partial least
squares structural equation modelling
(PLS-SEM) An emerging tool in business research. European Business Review, 26(2), 106-121.
Fornell, C.,
&Bookstein, F. L. (1982). Two structural equation models: LISREL and PLS applied
to consumer exit-voice theory. Journal of marketing research, 19(4),
440-452.
Fornell, C.,
&Larcker, D. F. (1981). Evaluating structural equation models with
unobservable variables and measurement error. Journal of Marketing
Research, 39-50.
Frenzen, H.,
Hansen, A. K., Krafft, M., Mantrala, M. K., & Schmidt, S. (2010).
Delegation of pricing authority to the sales force: An agency-theoretic
perspective of its determinants and impact on performance. International
Journal of Research in Marketing, 27(1), 58-68.
Gefen, D.,
Rigdon, E. E., & Straub, D. (2011). Editor's comments: an update and
extension to SEM guidelines for administrative and social science
research. Mis Quarterly, iii-xiv.
Hahn, C.,
Johnson, M.D., Herrmann, A. and Huber, F. (2002), “Capturing customer
heterogeneity using a finite mixture PLS approach”, Schmalenbach Business
Review, Vol. 54 No. 3, pp. 243-269.
Hair Jr, J. F.,
Hult, G. T. M., Ringle, C., &Sarstedt, M. (2016). A primer on
partial least squares structural equation modeling (PLS-SEM). Sage Publications.
Hair, Jr. F., Ringle, C. M., &Sarstedt, M. (2011). PLS-SEM: Indeed a
silver bullet. Journal of Marketing theory and Practice, 19(2), 139-152.
Hair, Jr. F.,
Sarstedt, M., Pieper, T. M., &Ringle, C. M. (2012). The use of partial
least squares structural equation modeling in strategic management research: a
review of past practices and recommendations for future applications. Long Range Planning, 45(5-6), 320-340.
Hair, Jr. F.,
Sarstedt, M., Ringle, C. M., & Mena, J. A. (2012). An assessment of the use
of partial least squares structural equation modeling in marketing
research. Journal of the academy of marketing science, 40(3),
414-433.
Hair, Jr, J. F., Sarstedt, M., Matthews, L. M., &Ringle, C. M.
(2016). Identifying and treating unobserved heterogeneity with FIMIX-PLS: part
I–method. European Business Review, 28(1), 63-76.
Henseler, J.,
Ringle, C. M., &Sinkovics, R. R. (2009). The use of partial least squares
path modeling in international marketing. In New challenges to international marketing (pp.
277-319). Emerald Group Publishing Limited.
Hoffmann, P.,
Schiele, H., &Krabbendam, K. (2013). Uncertainty, supply risk management
and their impact on performance. Journal of purchasing and supply
management, 19(3), 199-211.
Jöreskog, K. G.
(1970). A general method for analysis of covariance structures. Biometrika, 57(2),
239-251.
Jöreskog, K. G.
(1971). Simultaneous factor analysis in several populations. Psychometrika, 36(4), 409-426.
Kallunki, J. P.,
Laitinen, E. K., &Silvola, H. (2011). Impact of enterprise resource
planning systems on management control systems and firm performance. International
Journal of Accounting Information Systems, 12(1), 20-39.
Kamath, V.,
Rodrigues, L. L., & Desai, P. V. (2016). The significance of knowledge
management, innovation on firm performance in the Indian manufacturing sectors:
an empirical analysis. International Journal of Business Excellence, 9(2), 178-191.
Kaufmann, L.,
&Gaeckler, J. (2015). A structured review of partial least squares in
supply chain management research. Journal of Purchasing and Supply
Management, 21(4), 259-272.
Kesari, B.,
&Atulkar, S. (2016). Satisfaction of mall shoppers: A study on perceived
utilitarian and hedonic shopping values. Journal of Retailing and
Consumer Services, 31,
22-31.
Lee, L., Petter,
S., Fayard, D., & Robinson, S. (2011). On the use of partial least squares
path modeling in accounting research. International Journal of
Accounting Information Systems, 12(4), 305-328.
Marcoulides, G. A., & Saunders, C. (2006). Editor's comments: PLS: a
silver bullet?. MIS quarterly, iii-ix.
Matthews, L. M., Sarstedt, M., Hair, J. F., &Ringle, C. M. (2016).
Identifying and treating unobserved heterogeneity with FIMIX-PLS: Part II–A
case study. European Business
Review, 28(2), 208-224.
Nagarajan, V.,
Savitskie, K., Ranganathan, S., Sen, S., &Alexandrov, A. (2013). The effect
of environmental uncertainty, information quality, and collaborative logistics
on supply chain flexibility of small manufacturing firms in India. Asia Pacific Journal of Marketing and Logistics, 25(5), 784-802.
Nunnally, J. C.,
& Bernstein, I. H. (1994). Psychological
theory. New York, NY:
MacGraw-Hill.
Peng, D. X.,
& Lai, F. (2012). Using partial least squares in operations management
research: A practical guideline and summary of past research. Journal
of Operations Management, 30(6), 467-480.
Okazaki, S.,
Mueller, B., & Taylor, C. R. (2010). Measuring soft-sell versus hard-sell
advertising appeals. Journal of Advertising, 39(2), 5-20.
Peng, D. X.,
& Lai, F. (2012). Using partial least squares in operations management
research: A practical guideline and summary of past research. Journal
of Operations Management, 30(6), 467-480.
Reinartz, W.,
Haenlein, M., &Henseler, J. (2009). An empirical comparison of the efficacy
of covariance-based and variance-based SEM. International Journal of
research in Marketing, 26(4), 332-344.
Richter, N. F.,
Sinkovics, R. R., Ringle, C. M., &Schlaegel, C. (2016). A critical look at
the use of SEM in international business research. International
Marketing Review, 33(3), 376-404.
Ringle, C. M.,
Sarstedt, M., &Mooi, E. A. (2010). Response-based segmentation using finite
mixture partial least squares. In Data, 19-49.
Ringle, C. M., Sarstedt, M., & Straub, D. (2012). A critical look at
the use of PLS-SEM. MIS Quarterl, 36(1),
3-14.
Schwaiger, M., Sarstedt, M., & Taylor, C.
R. (2010). Art for the sake of the corporation: Audi, BMW Group,
DaimlerChrysler, Montblanc, Siemens, and Volkswagen help explore the effect of
sponsorship on corporate reputations. Journal of Advertising Research, 50(1),
77-90.
Seetharaman, A.,
Bajaj, S., Raj, J. R., &Saravanan, A. S. (2013). A Consumers' perception of
Wal-Mart in India. International Journal of Academic Research, 5(3).
Shanmugapriya,
S., & Subramanian, K. (2015). Structural equation model to investigate the
factors influencing quality performance in Indian construction projects. Sadhana, 40(6), 1975-1987.
SmartPLS.
(2018). PLS-SEM Compared with CB-SEM.
Retrieved on December 22, 2018 from smartpls.com:
https://www.smartpls.com/documentation/learn-pls-sem-and-smartpls/pls-sem-compared-with-cbsem.
Sosik, J. J., Kahai, S. S., &Piovoso, M. J. (2009). Silver bullet or
voodoo statistics? A primer for using the partial least squares data analytic
technique in group and organization research. Group & Organization
Management, 34(1), 5-36.
Stone, M.
(1974). Cross-validatory choice and assessment of statistical
predictions. Journal of the Royal Statistical Society. Series B (Methodological), 111-147.
Tenenhaus, M.
(2008), “Component-Based Structural Equation Modelling,” Total Quality
Management & Business Excellence,
19 (7–8), 871–886.
Venkatesh, V.,
Sykes, T. A., &Venkatraman, S. (2014). Understanding e‐Government portal use
in rural India: role of demographic and personality characteristics. Information
Systems Journal, 24(3), 249-269.
Wetzels, M.,
Odekerken-Schröder, G., & Van Oppen, C. (2009). Using PLS path modeling for
assessing hierarchical construct models: Guidelines and empirical
illustration. MIS quarterly, 177-195.
Whiteman, W. E.
(1998). Training and educating army
officers for the 21st century: Implications for the United States Military
Academy. Fort Belvoir, VA: Defense Technical Information Center.
Wold, H. (1974),
Causal flows with latent variables: partings of ways in the light of NIPALS
modelling, European Economic Review, 5(1) 67-86.
Wold, H. (1982).
Soft modeling: the basic design and some extensions. Systems under
indirect observation, 2, 343.
Wolfle, L. M.
(1980). Strategies of path analysis. American Educational Research
Journal, 17(2),
183-209.
Wong, K. K. K. (2013). Partial least squares structural equation modeling
(PLS-SEM) techniques using SmartPLS. Marketing Bulletin, 24(1), 1-32.
