A SIMULATION STUDY OF BOOTSTRAP CONFIDENCE INTERV ALS FOR THE LOCATION OF ASYMMETRIC AND HEAVY TAILED DISTRIBUTIONS

Abstract

Bootstrap methodology is a modern statistical tool which enables us makin g statistical inference when the sampling distribution of the estimator is not known. Although the underlying idea is the same in all bootstrap methods, one might come across so many variations in tthe literature. In this study, the coverage accuracy of four most commonly used bootstrap confidence interval methods was assessed for various asymmetr c and heavy tailed distributions with an exh austive Monte Carlo simulation. In most of the cases, it has been found that the coverage accuracy of bootstrap percentile method is close to nominal for robust estimators of location.