Re: "Confidence limits made easy: interval estimation using a substitution method".
Re: "Confidence limits made easy: interval estimation using a substitution method".
复制标题
回复：“置信极限变得容易：使用替代方法进行区间估计”。
DOI：
10.1093/oxfordjournals.aje.a009906
复制
发表时间：
1999
影响因子：
5
通讯作者：
R. Newcombe
中科院分区：
文献类型：
--
作者：
R. Newcombe
Daly（1）最近描述了他所谓的用于设置置信限的”替代方法”。虽然这种方法是众所周知的，并且适用于戴利论文中公式1和公式2中给出的相对风险，但他描述的人群归因分数（莱文归因风险）的应用是错误的，除非在某些特殊情况下。有两个原因可以解释为什么戴利的应用（用他的方程7-9来说明）在一般情况下是不正确的。首先，几乎总是需要调整混杂因素（至少是年龄和性别）。正如很久以前（2）和最近（3，4）重申的那样，将调整后的相对风险（RRJ）替换为莱文公式（Daly论文（1，pp. 785-6））将产生有偏估计和置信限，除非调整没有区别。通过使用Miettinen公式（5）对人口归因分数（估计值V）进行计算，可以非常简单地获得正确的点估计值。
Daly (1) recently described what he called the" substitution method" for setting confidence limits. While this method is well known and is correct for the application to relative risks given in formulas 1 and 2 of Daly's paper, the application he describes for the population attributable fraction (Levin's attributable risk) is erroneous, except in some special cases. There are two reasons why Daly's application, illustrated by his equations 7-9, is not correct in general. First, adjustment for confounders (at least age and sex) is almost always needed. As was shown long ago (2) and reiterated more recently (3, 4), substitution of an adjusted relative risk (RRJ into Levin's formula (equations 7-9 of Daly's paper (1, pp. 785-6)) will yield biased estimates and confidence limits, unless adjustment makes no difference. A correct point estimate can be obtained quite simply by using Miettinen's formula (5) for the population attributable fraction, estimate V. One can then back-calculate to the 95 percent