Comparative Estimation Study On Multiple Comparisons Procedures

Objectives:The comparison is a common problem in medical data processing. Comparing sevearl rate relies on multiple comparisons procedures. Alhtough over 5 methods had been proposed,there has been no consensus on which one is worth advocating at present in biosattisties. Therefore,exploring the advantages and disadvantages of each method by simulations is the main task in this research.Methods:There are 5 methods:Bonferroni,Du,SNK-Zar,Benjamini-Hochberg,Bootstrap. We systematically varied the number of hypotheses,the proportion of false null hypotheses,sample size in data simulations and examined the effect of each of these factors on FWER,FDR,Power.each is repeated 10,000 times.Results:Bonferroni used easily,but the result is to conservation;Du do not control the FDR at the desired 0.05 level;Bootstrap could used widely,but the result is to conservation;Conclusion :Unfortunately,however,a uniformly"best"method of these examined does not exist.we should select appropriate methods according to the actual demands.