The analysis of pair-matched case-control studies, a multivariate approach.

In matched case-control studies one frequently must consider more than one variable in the analysis and in this paper a log-linear model is presented to meet this objective. A conditional argument yields a method for making inferences on the parameters measuring the association between the variables and disease. The result is similar to the problem of fitting a Bradley-Terry model when one has paired comparisons. Two methods of obtaining maximum conditional likelihood estimates of the parameters are possible: (i) fitting a quasi-independence model in the usual log-linear models context and (ii) fitting a linear logistic model. The results are illustrated with examples.