Abstract: We present a class of projection method estimators for multiple hazard ratios in proportional hazards model, and develop an algorithm for the computation of the estimators and their estimated covariance matrices in this article. The estimators are generalizations of Mantel-Haenszel estimator, and it has been shown that they might be more efficient than the partial likelihood estimator for moderate and small sample sizes. The algorithm can accommodate tied failure times, and stratified models adjusting for the effects of confounding variables.