A MONTE-CARLO COMPARISON OF SOME BAYESIAN AND SAMPLING THEORY ESTIMATORS IN 2 HETEROSCEDASTIC ERROR MODELS

Abstract

Additive and multiplicative heteroscedastic error models are considered, and the relative efficiency of a number of Bayesian and sampling theory estimators, including some pre-test estimators,is investigated. The relative performance of the various estimators seems to depend more on sample size and the severity of the heteroscedasticity than it does on the underlying nature of the heteroscedasticity. In general the Bayesian estimators outperform the sampling theory estimators, and we make some useful recommendations concerning estimator choice.