A Bayesian Approach to Joint Modeling of Longitudinal Menstrual Cycle Length and the Probability of Pregnancy

• Activity Number: 452
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Kirsten J. Lum*+ and Rajeshwari Sundaram and Germaine M. Louis and Thomas Louis
• Companies: Johns Hopkins University/National Institute of Child Health and Human Development and Eunice Kennedy Shriver National Institute of Child Health and Human Development and Eunice Kennedy Shriver National Institute of Child Health and Human Development and U.S. Census Bureau/Johns Hopkins University
• Keywords: Bayesian ; Menstrual cycle ; Prospective pregnancy study ; Random effect ; Weight

Abstract

Menstrual cycle variation has received much attention for its potential role in a couple's probability of pregnancy. In order to examine the time-varying factors that may contribute to a couple's success in becoming pregnant, prospective pregnancy studies have been conducted in which couples record daily measurements of menstruation, sexual intercourse, and other factors until a successful pregnancy or censoring. In these studies, couples with a larger number of cycles provide more data on menstrual cycle characteristics and intercourse. Using a fully Bayesian model that relates menstrual cycle length to the probability of pregnancy, we study procedures for estimating population parameters that account for differences in the number of cycles and intercourse acts. Including one or more couple-specific random effects gives a weighting scheme that to a large degree accounts for differences in couple-specific information. We demonstrate this approach using data from the Longitudinal Investigation of Fertility and the Environment Study.