Effects on mantel-haenszel chi-square statistic of scores in two-way contingency tables

Abstract

Analysis of association between ordinal categorical variables has been widely studied especially in social sciences and medicine. It is known that ordinal scale has several advantages over nominal scale for researchers. MantelHaenszel (MH) chi-square statistic, which is known as the appropriate statistics to identify the association between ordinal categorical variables, takes into consideration the linear trend. In this study, integer, midrank, exponential, Van der Waerden, and joint scores were compared with a Monte Carlo simulation study to evaluate power of MH chi-square statistics. Although integer score has been used so far in literature, it is seen that Van der Waerden score is also preferable according to results.