The accuracy of two-sided confidence interval algorithm: an alternative of double bootstrap approach

Abstract

The basis idea of double bootstrap is using iteration procedure in the set sample of first bootstrap. The iteration seems to consume a period of time in completing all steps of double bootstrap algorithm and could delaying interested estimation such as accuracy estimation. To overcome this problem, an alternative of double bootstrap is proposed in this research. Instead of using iteration, generally, the procedure is to repeat the whole algorithm of single bootstrap in resulting a sample of double bootstrap. In terms of accuracy, this research constructs two double bootstrap intervals and estimates the coverage rate and its length. The procedure of double bootstrap and its accuracy estimation apply in simulation study. The result shows double bootstrap intervals have the same coverage with nominal rate and gives shorter length when comparing with single bootstrap intervals. Thus, the proposed double bootstrap has given an accuracy estimation and not complex to be used in simulation study.