A Comparison of Some Estimation Methods for Var-Covariance matrix in Big Data with Application

Abstract

The of the var-covariance matrix estimating is very important and represents a cornerstone in many statistical methods for several scientific fields, as many methods have emerged that are concerned with estimating the var-covariance matrix based on method of maximum likelihood. Nowadays these methods are classical methods and the process of estimating the covariance difficult with the increase in the number of variables under study. In this research, we used two methods to estimate the var-covariance matrix in big data in tow methods . first is Banding Estimator BE and the Tapering Estimator TE. Simulation study were used, as well as real data for living standards obtained from The Iraqi Ministry of Planning, the percentage improvement in average loss (PRIAL) criterion was used, as it was concluded that the BE Banding Estimator works better than the TE Tapering Estimator under big data conditions.