成组设计两样本率非劣效性检验方法的比较研究

Comparative Study of Noninferiority Test on Two Sample Rates of Group Design

目的对非劣效性检验的不同方法进行比较研究。方法通过计算正态分布法和精确检验的Ⅰ类错误与检验效能,对两种方法做出比较与评价。结果小样本或两组样本例数相差悬殊时,正态分布法的Ⅰ类错误可能会超过名义上的检验水准;多数情况下正态分布法的检验效能高于精确检验,但是在小样本或两组样本例数相差悬殊时,会出现精确检验的检验效能较高的情况。结论定性资料的非劣效性检验在大样本时应该选择正态分布法,但是在小样本或两组样本例数相差悬殊时,需要考虑采用精确检验。

Objective To compare the different methods of noninferiority test. Methods We evaluated the type I error and power of test between the methods of normal distribution test and exact test. Results Type I error may exceed the nominal size of test when disparity existed between the two samples or in small samples. Usually ,the power of test of normal distribution test was superior to that of exact test, but the power of test of exact test was superior when disparity existed between the two sam- ples or in small samples. Conclusion Normal distribution test was eligi- ble for noninferiority test of large samples of qualitative data. However, ex- act test may be used when disparity existed between the two samples or in small samples.