Modeling stress: a methodological review

Covariance structure modeling (CSM) is increasingly being utilized in stress research. For this reason, a methodological review was undertaken to assess how well applied researchers have recognized the shortcomings and recommendations commonly espoused in the mathematical and statistical CSM literature. In reviewing a broad cross section of research (n = 50), many problems were noted. For example, data sets were rarely checked for nonnormality even though conventional CSM estimation procedures (e.g., maximum likelihood) are based on the assumption that the data come from a multivariate normal distribution, equivalent models were neither evaluated nor even acknowledged, and the majority of favored models were modified based on empirical rather than theoretical considerations. The results suggest that applied stress researchers have inadequate knowledge of the technical CSM literature, which has potential ramifications for theory development.