Abstract |
The analysis of case-control data using logistic regression involves making some critical model assumptions. In the analysis of matched data for which the values of the matching variables are unavailable, use of logistic regression presupposes that odds ratios are homogeneous over the levels of the matching variables; this is done since logistic regression cannot, at present, provide a way of testing for the homogeneity of odds ratios. In the model, it is possible, as in logistic regression, to estimate odds ratios, check the multiplicativity of effects, or test for the significance of association. In addition, and unlike logistic regression, the model enables one to test for the homogeneity of the odds ratios over the levels of the matching variables even, when explicit information is not available on the matching variables. |