Abstract: Conditional logistic regression model was developed in recent years. We suggest a conditional mean method to select important variables in matohed data studies to overcome some difficulties pointed out in our paper. Besides, we prove the equivalentce between the conditional mean method and the additional information method, but the former gives indication of the direetions of the variable influence between populations, and the distributions of the mashed sets aren’t necessary identical. The computing procedure and properties of the oonditional mean method are shown, an example of matched lung cancer data is presented.