A Semiparametric Mixture Model for Stratified Survival Data, with an Application to the Study of Transition to Teenage Motherhood in South Africa
Samuel O.M. Manda, South Africa Medical Research Council
Renate Meyer, University of Auckland
Bo Cai, University of South Carolina
Low rates of teenage pregnancy contribute to the improvement of maternal and child health and the eradication of poverty, part of the Millennium Development Goals. Despite recent declines worldwide, the current levels of teenage pregnancy in South Africa and many developing countries are still much higher compared to the developed countries. We employ time-to-event models to investigate important biodemographic, social-economic and cultural determinants of early first child birth in South Africa using data from a national representative stratified cluster sample of women between the ages of 15 and 59 years. The stratum-specific baseline transitions to early motherhood are flexibly modelled with non-parametric techniques based on a mixture of distributions. We consider mixtures of beta and triangular distributions which are fitted to the data within a Bayesian framework. The results are compared to those obtained from fitting a standard parametric Weibull model for the stratum-specific baseline hazard functions.
Presented in Poster Session 1