Longitudinal epidemiologic studies of menstrual and reproductive function are more informative if one can identify day of ovulation. We previously developed a method for estimating day of ovulation that is feasible for epidemiologic studies. The method relies on the relative concentrations of estrogen and progesterone metabolites in daily first-morning urine specimens and does not require creatinine adjustment. This paper describes results of applying this method to a large study with 724 menstrual cycles from 217 women. The method estimated a credible day of ovulation in 88% of cycles. Missing data accounted for most of the failures. When we excluded anovulatory cycles (1%) and cycles with missing data, the method estimated a day of ovulation in 97% of cycles. Variance in luteal phase length was small for our sample, suggesting that this method of identifying a day of ovulation introduces no more measurement error than when day of ovulation is determined by plasma luteinizing hormone (LH), the standard clinical method.