Ranked simulated resampling: a more efficient and accurate resampling approximations for bootstrap inference

Abstract

Since its invention, Efron’s bootstrap resampling approach has changed all the aspects of statistical inference, which has become the default framework whenever the classical inference approaches are not feasible. This paper introduces a new, more accurate, and efficient resampling approach, namely, the ranked simulated resampling approach. We show that, analytically and computationally, it is more efficient and precise than Efron’s uniform bootstrap resampling approach. We provide simulation studies and real data applications to support the comparison between the ranked simulated resampling approach and Efron’s uniform bootstrap resampling approach.