Bias reduction in the equicorrelated multivariate normal

In the multivariate normal model, the maximum likelihood estimates can be highly inaccurate with small sample size, or in presence of many covariates. The variance and correlation may result in substantial bias and therefore compromise the inferential conclusions.The paper focuses on the equicorrelated normal model and uses the mean and median bias reduction methods to improve the accuracy of inference. The properties of the resulting estimators are assessed through extensive simulation studies and one application.