On logit confidence intervals for the odds ratio with small samples

被引：87

作者：

Agresti, A ^{[1
]}

机构：

[1] Univ Florida, Dept Stat, Gainesville, FL 32611 USA

来源：

BIOMETRICS | 1999年 / 55卷 / 02期

关键词：

bayes estimate; chi-squared test; contingency table; delta method; exact inference; two-by-two table;

D O I：

10.1111/j.0006-341X.1999.00597.x

中图分类号：

Q [生物科学];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

Unless the true association is very strong, simple large-sample confidence intervals for the odds ratio based on the delta method perform well even for small samples. Such intervals include the Woolf logit interval and the related Cart interval based on adding .5 before computing the log odds ratio estimate and its standard error. The Cart interval smooths the observed counts toward the model of equiprobability, but one obtains better coverage probabilities by smoothing toward the independence model and by extending the interval in the appropriate direction when a cell count is zero.

引用

页码：597 / 602

页数：6

共 18 条

[1] Approximate is better than "exact" for interval estimation of binomial proportions [J].