Analysis of recurrent failure times data: should the baseline hazard be stratified