On the treatment of correlation structures as covariance structures (Q918092)

This paper investigates conditions under which methods intended for the analysis of covariance structures result in correct statistical conclusions when employed for the analysis of correlation structures. The authors give necessary and sufficient conditions for the asymptotic chi- squaredness of the minimum discrepancy function (MDF) test statistic and the asymptotic efficiency of the corresponding MDF estimators. Examples of correlation structure satisfying these conditions are considered. Alternative discrepancy functions, which are always appropriate for the analysis of correlation structures under normality assumptions, are also discussed.