Regression analysis of matched case-control data.

A regression methodology is outlined for the logistic analysis of matched pair case-control data. This weighted least squares technique is applicable to the problem of analyzing multiple, but categorical, risk factors. Estimates of odds ratios and statistics for hypothesis testing are directly produced with the procedure. The proposed methodology may be viewed as an alternative to that described by Holford et al. (Am J Epidemiol 1978;107:245-56). The required computations may be performed with standard programs for least squares categorical data analysis.