Comparison of different methods for adjusting for confounding in a Cox regression using simulated data in stata