Application 9: Testing for Invariant Causal Structure Across Calibration and Validation Samples

In chapter 4, I highlighted several problematic aspects of post hoc model fitting in structural equation modeling. Indeed, so common is this practice—and so frequently is it conducted with little to no regard for the substantive meaning-fulness of the respecified models—that concerned researchers have provided a variety of different means by which such models can be tested more stringently (see, e.g., Anderson & Gerbing, 1988; Cudeck & Henly, 1991; MacCallum, 1995; MacCallum et al., 1992, MacCallum et al., 1993). In chapter 4, I noted that one approach to addressing problems associated with post hoc model fitting is to apply some type of cross-validation tedchnique; one such strategy was illustrated in that chapter.