Eunice Ampah
PhD student, Department of Statistics
Western Michigan University
Abstract:
In epidemiology and medical research,
data is often summarized in a 2×2 contingency table.
The table represents cross-classification between
disease outcome and exposure to a risk factor.
In the presence of confounders, the tables are
stratified according to values of the confounding variable,
and association is tested using the Cochran-Mantel-Haenszel statistic.
We propose a testing procedure based on log-odds ratios as an alternative or
supplement to the CMH statistic when investigating association across strata.
Performance of the proposed test is compared to the CMH statistic
under various null and alternative distributional patterns
by evaluating size and power thorugh a simulation study.
We examine conditions under which
the proposed test outperforms the CMH statistic.
---
All statistics graduate students are expected to attend.