PARSIMONY AND MODEL EVALUATION

Marsh and Hau (1996) argued that certain models should not be penalize d for having low parsimony because an appropriate model for the data m ay require estimating more parameters. Mulaik argues that Marsh and Ha u misunderstand the concept of parsimony, particularly its role in tes ting a hypothesis about an incompletely specified model to establish i ts objective validity, More parsimonious models represent more complet e hypotheses having more,cays of being tested and possibly being disco nfirmed. Mulaik also shows that even within the context of the models used in Marsh and Hau's examples, there are much more parsimonious ver sions of those models that could have been hypothesized and tested, wi th good fit.