Testing the Differential Impact of Structural Paths in PLS Analysis: A Bootstrapping Approach

Abstract

Researchers are often interested in examining the relative impact of PLS structural paths. As such, this paper focuses on how one assesses the impact of various antecedent constructs to a particular endogenous construct. In a survey of recent papers in the field of Information Systems employing structural equation modeling, discussion of differences or conversely equivalency of paths were typically made without any statistical tests. In a few cases, a traditional t-test was used. This paper begins with a didactic presentation of how such t-tests are estimated followed with introducing an alternative bootstrapping approach. Results from both empirical and simulated data show different conclusions are made between these two approaches. In particular, we show that under data conditions of high kurtosis, bootstrapping is less likely to commit a Type I error of stating substantial differences among paths when none exist.