多重相关检验中错误发现率的控制算法

假设检验问题是通过比较p值和置信水平α的值来决定是否拒绝对应的假设,pα时我们拒绝原假设。随着试验次数的增加,在所有满足零假设为真的p值集合中,数值较小的p值存在的可能性会增加,从而使得做出错误的判断,p值调整算法可以针对多重检验有效地缓解这类问题。本文讨论多重检验的p值调整算法的功效,模拟基因序列进行多重检验分析,产生2000个模拟量(即基因数量),重复实验1000次,得到1000组p值。对每组p值使用相应的调整算法得到新的p值,比较每个算法功效的优劣。模拟结果显示在我们所选用的5种p值调整算法中,q值方法(Storey 2003)能很好地控制错误发现率(FDR)的大小,同时具有更高的功效值。

RESEARCH ON ALGORITHM OF FALSE DISCOVERY RATE CONTROL IN MULTIPLE DEPENDENT TESTS

Hypothesis testing problem determines whether to reject the corresponding hypothesis through comparing p values with confidence level. A smaller p-value than the -level of the test signifies a statistically significant test. As the number of tests increases, the chance of observing some small p-values is very high even when all null hypotheses are true. Consequently, we make wrong conclusions on the hypotheses. Adjustment of p-values can effectively alleviate this problem. A simulation study with several methods was carried out in multiple hypothesis testing. Simulations generate 2000 analogue. The experiment was repeated 1000 times, 1000 sets of p value are obtained. Use the adjustment methods to access the new p-value, and then compare the advantages and disadvantages of each algorithm. The correlation between genes expression is in the covariance matrix. Simulation results show that among five algorithms selected, q value method can be efficient to control the false discovery rate size, also with higher power.