A comparison of maximum likelihood and Jewell's estimators of the odds ratio and relative risk in single 2 x 2 tables.

This paper compares the maximum likelihood estimators of the odds ratio and relative risk in 2 x 2 tables with less biased estimators recently proposed by Jewell for small samples. In addition to bias, location in the confidence interval, and relative likelihood are calculated for ranges of sample sizes, probabilities (of exposure or disease) and true values of the odds ratio and relative risk. Jewell's estimators perform poorly from the perspectives of confidence intervals and likelihood. Given these and other problems, continued use of the maximum likelihood estimators is recommended.