Testing the normality assumption in multivariate simultaneous limited dependent variable models

Abstract

A diagnostic for multivariate non-normality expressed as a multivariate Edgeworth distribution truncated at the fourth order is presented for a multivariate simultaneous equation system in normally distributed latent variables subject to a general censoring scheme. The information matrix test statistic also presented tests for multivariate non-normality and particular forms of multivariate heteroskedasticity and heterocliticity. Alternative C(α) tests are also suggested. © 1987.