Research on null hypothesis of statistical test;

This paper points out the problems about using statistical hypothesis test, analyses the relation between two types of errors and proposes the Bayes method to the persons when they have no enough knowledge about the null hypothesis.

First of all,the designing principles of null hypothesis are discussed,then the limits of significance level and p-value method are introduced.

In hypothesis test,there exists certain kind of phenomenon: for the given sample,if the original hypothesis is different,the statistic inference conclusion may be different.

In this article we will analyze the problems, from the point of dialectics, caused by changing original hypothesis with choice hypothesis in unilateral hypothesis - test.

The first to solve and the most difficult problem in hypothesis is the pose of original hypothesis,which is signed H_0.

According to the basic principle of hypothesis testing,three kinds of one sided hypothesis testing with the same reject region at the same significant level α are given and how to choose primary hypothesis and alternative hypothesis at the actual application and correlation are also pointed out.

This paper discuss the relationship between two types of errors in statistical hypothesis , and points out choosing the primary hypothesis and the significant level in the actual application .