Empirical comparison of ML and UMVU estimators of the generalized variance for some normal stable Tweedie models: a simulation study

Abstract

This paper discuss a comparison of the maximum likelihood (ML) estimator and the uniformly minimum variance unbiased (UMVU) es- timator of generalized variance for some normal stable Tweedie models through simulation study. We describe the estimation of some particular cases of multivariate NST models, i.e. normal gamma, normal Poisson and normal invers-Gaussian. The result shows that UMVU method pro- duces better estimations than ML method on small samples and they both produce similar estimations on large samples.