论文部分内容阅读
在医学研究中,对于一个研究对象经常需要观察两种及两种以上的指标。例如,测量血压要观察舒张压和收缩压两个指标;研究青少年的生长发育状况则需观察身高、体重和胸围等指标。对于这类资料的均数假设检验(即均数向量的假设检验),人们习惯于用t检验(或u检验)的方法对每个指标分别作均数的假设检验。这种方法虽然简便,但由于作为一个整体的各个指标是互相联系的,当各个指标相互独立时,指标愈多犯Ⅰ型错误(即假阳性)的概率愈大;当各指标相关时,则不易发现差异。
In medical research, it is often necessary to observe two or more indicators for a subject. For example, measuring blood pressure to observe the two indicators of diastolic and systolic pressure; study the growth and development of young people need to observe height, weight and chest circumference and other indicators. For the mean-number hypothesis test of this type of data (ie, the hypothesis test of the mean vector), people are accustomed to using the t-test (or u-test) to perform a hypothesis test on the mean of each indicator. Although this method is simple, but the indicators as a whole are interconnected, when the indicators are independent of each other, the more the indicators are more likely to make Type I errors (ie false positives); when the indicators are related, then It is not easy to find the difference.