Mickey, Ruth M.; Elashoff, Robert M. A generalization of the Mantel-Haenszel estimator of partial association for 2\(\times J\times K\) tables. (English) Zbl 0617.62059 Biometrics 41, 623-635 (1985). A generalization of the Mantel-Haenszel estimator of a common odds ratio for a series of \(2\times 2\) tables is proposed for the estimation of log linear partial association parameters for a series of \(2\times J\) \((J>2)\) tables. The assumption of no three-way interaction among row, column, and stratifying variables is made. Two approximate covariance estimators are presented. Small-sample properties of these estimators are investigated by simulation studies. The results indicate that use of the proposed partial association technique produces estimates with negligible bias. Both covariance estimation techniques provide confidence intervals for functions of partial association parameters, for example, odds ratios or adjusted rates. A test of no three-way interaction results from these procedures. Cited in 1 ReviewCited in 2 Documents MSC: 62H17 Contingency tables 62H12 Estimation in multivariate analysis Keywords:Mantel-Haenszel estimator; odds ratio; estimation of log linear partial association parameters; three-way interaction; approximate covariance estimators; Small-sample properties; simulation studies; confidence intervals PDFBibTeX XMLCite \textit{R. M. Mickey} and \textit{R. M. Elashoff}, Biometrics 41, 623--635 (1985; Zbl 0617.62059) Full Text: DOI