Analysis of structural equation models with censored data

A two-stage estimation procedure for analyses of structural equation models with censored data is developed in this paper. The first stage involves the partition maximum likelihood estimation of the unstructured covariance matrix. It is shown that the joint asymptotic distribution of these parameter estimates is multivariate normal. An analytic expression for the asymptotic covariance matrix is derived. In the second stage, the structural parameters are estimated by means of the generalized least squares approach with an appropriately specified weight matrix. A simulation study is conducted to study the performance of the proposed approach and to compare it with that of LISCOMP.