A comparison of four methods of calculating standard errors of maximum‐likelihood estimates in the analysis of covariance structure

Four methods of calculating standard errors (SEs) in the analysis of covariance structure using normal theory maximum‐likelihood estimation, namely LISREL SEs, SEs derived from the exact Hessian, SEs derived from the finite difference approximation to the Hessian using exact gradients, and SEs derived from the update of the Hessian obtained during quasi‐Newton optimization, are compared. Two comparisons are made. The first is based on 30 data sets simulated according to an orthogonal common factor model. The second is based on the analysis of a well known data set concerning the stability of alienation. The finite difference and exact SEs are identical to at least four places of decimals. Little difference is observed between the remaining SEs. 1991 The British Psychological Society