Inaccuracy of regression results in replacing bivariate correlations
Inaccuracy of regression results in replacing bivariate correlations
复制标题
DOI：
10.1002/jrsm.1126
复制
发表时间：
2015-03-01
影响因子：
9.8
通讯作者：
Aloe, Ariel M.
中科院分区：
文献类型：
--
作者：
Aloe, Ariel M.
这篇手稿考虑了双变量相关性和从回归结果估计的几个关联指数之间的差异。这些指数可以从初步研究中通常报告的结果中估计出来。近年来，许多进行荟萃分析的研究人员使用这些指标来代替或与双变量相关一起使用。我举例说明了这些指数之间的差异和双变量相关性。我证明了这些指数作为双变量效应的替代物是不准确的。因此，我建议停止使用这些指数和部分效应大小来替代双变量相关性。版权所有(C)2014 John Wiley&Sons，Ltd.
This manuscript considers discrepancies between the bivariate correlation and several indices of association estimated from regression results. These indices can be estimated from results typically reported in primary studies. In recent years, many researchers conducting meta-analyses have used these indices in place of, or together with, the bivariate correlation. I illustrate the differences among these indices and the bivariate correlation. I demonstrate the inaccuracy of these indices as replacements for bivariate effects. Thus, I recommend discontinuing the use of these indices and partial effect sizes as replacement for the bivariate correlation. Copyright (c) 2014 John Wiley & Sons, Ltd.