Rethinking some of the rethinking of partial least squares

PurposePartial least squares structural equation modeling (PLS-SEM) is an important statistical technique in the toolbox of methods that researchers in marketing and other social sciences disciplines frequently use in their empirical analyses. The purpose of this paper is to shed light on several misconceptions that have emerged as a result of the proposed “new guidelines” for PLS-SEM. The authors discuss various aspects related to current debates on when or when not to use PLS-SEM, and which model evaluation metrics to apply. In addition, this paper summarizes several important methodological extensions of PLS-SEM researchers can use to improve the quality of their analyses, results and findings.Design/methodology/approachThe paper merges literature from various disciplines, including marketing, strategic management, information systems, accounting and statistics, to present a state-of-the-art review of PLS-SEM. Based on these findings, the paper offers a point of orientation on how to consider and apply these latest developments when executing or assessing PLS-SEM-based research.FindingsThis paper offers guidance regarding situations that favor the use of PLS-SEM and discusses the need to consider certain model evaluation metrics. It also summarizes how to deal with endogeneity in PLS-SEM, and critically comments on the recent proposal to adjust PLS-SEM estimates to mimic common factor models that are the foundation of covariance-based SEM. Finally, this paper opposes characterizing common concepts and practices of PLS-SEM as “out-of-date” without providing well-substantiated alternatives and solutions.Research limitations/implicationsThe paper paves the way for future discussions and suggests a way forward to reach consensus regarding situations that favor PLS-SEM use and its application.Practical implicationsThis paper offers guidance on how to consider the latest methodological developments when executing or assessing PLS-SEM-based research.Originality/valueThis paper complements recently proposed “new guidelines” with the aim of offering a counter perspective on some strong claims made in the latest literature on PLS-SEM. It also clarifies some misconceptions regarding the application of PLS-SEM.

WorseThanMeasurementError

Previous research and methodological advice has focused on the importance of accounting for measurement error in psychological data. That perspective assumes that psychological variables conform to a common factor model, such that they consist of construct variance plus error. In this paper, we explore what happens when a set of items that are not generated from a common factor construct model are nonetheless modeled as reflecting a common factor. Through a series of hypothetical examples and an empirical re-analysis, we show that (1) common factor models tend to produce extremely biased and highly variable structural parameter estimates when the population model is not a common factor model; (2) model fit is a poor indicator of the degree of bias; and (3) composite models are sometimes more reliable than common factor models under alternative measurement structures, though they also lead to unacceptably bad solutions in some cases.

Genetic Association in Multivariate Phenotypic Data: Power in Five Models

This article concerns the power of various data analytic strategies to detect the effect of a single genetic variant (GV) in multivariate data. We simulated exactly fitting monozygotic and dizygotic phenotypic data according to single and two common factor models, and simplex models. We calculated the power to detect the GV in twin 1 data in an ANOVA of phenotypic sum scores, in a MANOVA, and in exploratory factor analysis (EFA), in which the common factors are regressed on the genetic variant. We also report power in the full twin model, and power of the single phenotype ANOVA. The results indicate that (1) if the GV affects all phenotypes, the sum score ANOVA and the EFA are most powerful, while the MANOVA is less powerful. Increasing phenotypic correlations further decreases the power of the MANOVA; and (2) if the GV affects only a subset of the phenotypes, the EFA or the MANOVA are most powerful, while sum score ANOVA is less powerful. In this case, an increase in phenotypic correlations may enhance the power of MANOVA and EFA. If the effect of the GV is modeled directly on the phenotypes in the EFA, the power of the EFA is approximately equal to the power of the MANOVA.

On the Mining of Cointegrated Econometric Models

The authors describe a process of extracting a cointegrated model from a database. An important part of the process is a model generator that automatically searches for cointegrated models and orders them according to an information criterion. They build and test a non-heuristic model generator that mines for common factor models, a special kind of cointegrated models. An outlook on potential future developments is given.