This paper discusses the need for multiple comparison and cross-validation in evaluating structural equation models from the perspectives of philosophy of science, statistics, and methodology. After the issues associated with a single model test are examined, a comparison of competing models is suggested in its place. Theoretical arguments are also made for the use of cross-validation in comparing alternative models.

Over the last decade, structural equation models have been used by many marketing researchers (e.g., Fornell 1987). Nevertheless, relatively few studies have addressed the question of how structural equation models should be evaluated (cf. Bagozzi 1981; Fornell and Larcker 1981). Furthermore, many of these studies have focused mainly on the statistical problems with current modeling procedures (Bagozzi and Yi 1988). The purpose of this paper is to investigate the issues in evaluating structural equation models from a different perspective based on the principles of scientific inferences.