Research On The Proportion Of Null Hypothesis In Multiple Hypothesis Testing

In bioinformatics, with the development of genetic technology, mass data has been coming into being, the typical multiple hypothesis testing is how to test the difference of mass data.We mainly do some research on multiple hypothesis testing in this paper. Firstly,we discuss whether the proportion of null hypothesis could be estimated, and we give the condition of the identification of the proportion of null hypothesis. Above all, we put forward a new method to estimate the proportion of null hypothesis, and we have proved that the estimation value by this new method converge to the true value. Under certain conditions, we could give the convergence rate.In addition, by simulation, we find that the method we present in the condition that null hypothesis is relatively close to alternative hypothesis, and the proportion of true null hypothesis is relatively small is better than the method of Storey, J.. namely, under the same testing levels, the power and the false rates of our method are better than the method of Storey, J..In the first chapter of this paper,we list the preparation knowledge of multiple hypothesis testing, and discuss the identification of the proportion of null hypothesis in the second chapter. In the third chapter, we discuss some methods of the estimation of the null hypothesis proportion, and we present Zhijiefa . In the four chapter, we mainly weigh these methods according to some typical examples.