TESTS FOR HOMOGENEITY OF EFFECT IN AN EPIDEMIOLOGIC INVESTIGATION
TESTS FOR HOMOGENEITY OF EFFECT IN AN EPIDEMIOLOGIC INVESTIGATION
复制标题
DOI：
10.1093/oxfordjournals.aje.a112441
复制
发表时间：
1977-01-01
影响因子：
5
通讯作者：
BYAR, DP
中科院分区：
文献类型：
--
作者：
MANTEL, N;BROWN, C;BYAR, DP
在人口的不同阶层中，一个诱发疾病的因素的影响的同质性检验应与预期同质性的尺度有关。一般来说，它是预期同质性的优势比的乘法效应。M.此缩放中的同质性Zelen无效。J.L.的测试Fleiss标准化差异的均匀性与Zelen检验相同，在任何情况下都无效，并且不对应于任何缩放。给出的例子包括1，其中2项研究的结果显然是不均匀的测试，即使他们实际上给出了相同的结果。同质性检验应谨慎进行，应考虑如何使效应同质化，而不是试图证明异质性的存在。
A test for homogeneity of the effect of a factor in inducing disease in the various strata of a population should relate to the scaling in which homogeneity is anticipated. Popularly, it is the multiplicative effect on the odds ratio in which homogeneity is anticipated. An asymptotic test by M. Zelen for homogeneity in this scaling is invalid. A test by J.L. Fleiss for homogeneity of the standardized difference is the same as the Zelen test, invalid in any case, and does not correspond to any scaling. The examples given include 1 in which results of 2 studies are apparently inhomogeneous by the test even though they give effectively identical results. Tests for homogeneity should be conducted with care, and thought should be given to how effects might be made homogeneous rather than to trying to demonstrate the existence of heterogeneity.