论文部分内容阅读
问题的提出医学研究中的计数资料常用列联表来表达,并通过假设检验来推论两样本的总体是否相同,通常以α=0.05(或0.01)作为检验水准。检验结果若拒绝检验假设,此时第一类错误的概率α≤0.05,第一类错误的概率即假阳性概率。医学科研文献中已广泛使用这种处理方法,某些文献中使用假设检验多达10余次。多次假设检验时,假阳性情况将是如何呢?这个问题比较棘手。上述假阳性概率α=0.05(或0.01)是指一次假设检验而言。根据数理统计原理,当医学研究中需进行n次假设检验时,这种假阳性概率就会随之增高,用公式表示为1—(1—α)~n。譬如一个研究中用了5次假设检验,各次检验皆以α=0.05为检验水准,则假阳性概率为1—(1—0.05)~5=0.2262,10次检验为1—(1—0.05)~(10)=0.4013,显然假
Question put forward in the medical research count data commonly used contingency table to express, and hypothesis test to infer the two samples are the same, usually α = 0.05 (or 0.01) as the test level. If the test results reject the hypothesis test, then the probability of the first kind of error is a≤0.05, the probability of the first kind of error is the false positive probability. This approach has been widely used in medical research literature, with some hypotheses tested in some literature up to more than 10 times. How many false positives will be the case when multiple hypotheses are tested? This is a tricky question. The above-mentioned false-positive probability α = 0.05 (or 0.01) refers to a hypothesis test. According to the principles of mathematical statistics, when the medical research needs n hypothesis tests, the false positive probability increases with the formula, which is expressed as 1- (1-α) ~ n. For example, in one study, five hypothesis tests were used. Each test was conducted with α = 0.05 as the test level, and the probability of false positives was 1- (1-0.05) ~ 5 = 0.2262. The test times were 1- (1-0.05 ) ~ (10) = 0.4013, apparently false