Reliability of results by matching methods for analytical epidemiology

点击翻译按钮获取中文摘要

In an analytical epidemiology a purpose of a matching method is to remove confounding variables. It is usually carried out in design of the study. It has been reported that the significance of it is validity and efficiency. When the effect of the confounding variables on the factors and the disease which are the subjects of the study is remarkable, the efficiency of it is significant. It is, however, mentioned that the reliability of it has not been thoroughly studied. Age factor affects incideces and mortality rates of diseases very remarkablly. The study on the caliper matching for the univariate continuous confounding factor in a cohort study was carried out statistically and numerically as an example of age.We revealed that (1) when the incience increasing rate of the disease by age or when the probability density function increasing rate of a control group by age is different from 1, the the expected incidence ratio of the disease for a caliper matched control is different from 1, (2) when the tolerance of the caliper matching is great, the ratio is different from 1.A caliper matched control is usually selected from the control group whose age is within the age of the case or the exposed suject <-!+> <THETA> years. Cancer incidence of older group is greater than that of younger. When such a matching method for cancer epidemiology is adopted, the expected incidence ratio of the cancer for a caliper matched control is greater than 1. We indicate that the caliper matching method proposed by us can avoid such a situstion.