Sample size determination method for credibility evaluation based on statistical power analysis

摘要

In this manuscript, we investigate the problem of sample size determination (SSD) method for rational prior credibility evaluation. First, in allusion to several existing credibility evaluation methods, risk of prior credibility measurement with data distribution difference is analyzed in small sample size conditions. Then, the relationships between and among significance level, effect size, sample size and statistical power are confirmed for data consistency test, which leads to the proposed power analysis based SSD method in credibility evaluation. After that, common used power analysis methods in hit accuracy test of weapon system, as those of t test for consistency of mean and F test for homogeneity of variance, are studied for corresponding SSD method. Finally, power analysis procedure and SSD method in credibility evaluation are illustrated with application examples, comparative analysis for superior rate of credibility weighted Bayesian assessment under small sample size condition and required statistical power condition is made, which proves the rationality of power analysis based SSD method in prior credibility evaluation.