ABSTRACT

The goal of the current chapter is to examine alternative model specification (AMS) practices as applied in covariance structure modeling (CSM). CSM is our general term referring to tests of confirmatory factor analysis (CFA) and/or structural equation models (SEM). Namely, the concern of this chapter is with the practices per se underlying AMS and, in particular, with the fact that AMS is seldom undertaken in light of long-standing and overwhelming advice to do so. At the risk of oversimplifying the issue, the concern is best illustrated by examining the extreme views on the issue. At one extreme, AMS advocates claim that it should be practiced doctrine and a regular feature in each and every CSM application (Greenwald, Leippe, Pratkanis, & Baumgardner, 1986; MacCallum & Austin, 2000; Reichardt, 2002). From this perspective, the presumption is that very little AMS is undertaken when it should be, and as a consequence, the validity of the results from the CSM is questionable. This consequence is due to researchers specifying only one model when alternative models using the same variables exist that are equivalent or better than the focal model (Boomsma, 2000; MacCallum, Wegener, Uchino, & Fabrigar, 1993; Williams, Bozdogan, & Aiman-Smith, 1996). At the other extreme of the AMS issue are the researchers who thoroughly undertake their theoretical “homework” and, as such, can anchor the paths between latent variables and/or items to underlying constructs solidly to conceptual arguments and frameworks. From their perspective, there is simply

no conceptual rationale to present an alternative specification to the focal model.