A Post-Positivist Answering Back.: Part 2: A Demo in R of the Importance of Enabling Replication in PLS and LISREL

Abstract

In Part 1 the argument was made for the core Positivist principle of enabling falsification, specifically that data, or at least correlation or covariance matrices, should be made public so that others can attempt to falsify at least the statistical analyses. Doing so could provide a semblance of the direction of what might constitute the desired Positivist aspects of intellectual integrity in science: making your claims and putting your data in the public domain so others may put its propositions to the test and try to falsify or improve on them. Part 2 demonstrates the importance of such disclosure. The demo begins with replicating the model in Structural Equation Modeling and Regression: Guidelines for Research Practice (Gefen et al., 2000) in PLS and CBSEM R packages, producing equivalent results as the original paper. Showing the point about the need to have the data in the public domain, a set of incorrectly specified models on the same data are then run. Both PLS and CBSEM converge and produce plausibly believable results if the data were not available to test alternative models, opening the possibility of pulling the wool over readers' eyes if such a correlation or covariance matrix is not provided.