Conditions for Confirmatory Analysis and Causal Inference.

摘要

Confirmatory analysis refers to a family of analytic procedures designed to evaluate the utility of causal hypotheses and to support inferences regarding causality among naturally occurring events derived from field studies. These analytic procedures are gaining rapid exposure in psychology, and include techniques such as confirmatory factor analysis, linear structural relations, path analysis, structural equations, and time series. This report specifies and discusses the conditions that justify the use of confirmatory analysis to evaluate causal hypotheses and to support causal inferences. The conditions include the need for strong theory, specification of causal order and direction, meaningful operationalization of variables, and statistical demonstration of goodness of fit between theoretical models and empirical data. (Author)