|
Activity Number:
|
144
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306096 |
|
Title:
|
Joint Modeling of Birthweight and Gestational Age
|
|
Author(s):
|
Betsy Enstrom*+ and Alan E. Gelfand and Geeta K. Swamy and Marie L. Miranda
|
|
Companies:
|
Duke University and Duke University and Duke University Medical Center and Duke University
|
|
Address:
|
5500 Bradford Pear Court, Raleigh, NC, 27606,
|
|
Keywords:
|
mixture model ; MCMC algorithm ; EM algorithm ; cumulative logit
|
|
Abstract:
|
Modeling birthweight (BW) and gestational age (GA) have received much discussion in the literature and it's linkage to social, environmental, and genetic risk factors. A clear definition of these two birth cohorts is beneficial to many fields of science. We propose joint modeling with the objective of better characterization of adverse birth outcomes. The data set consists of 107045 live births as recorded in the North Carolina Detailed Birth Record for 2003, we develop the model specifications in a hierarchical Bayesian framework as well as a likelihoodist setting. A MCMC and EM algorithm are illustrated for the Bayesian and frequentist approach, respectively. Practical issues that arise when fitting these types of models are discussed including the number of mixture components, convergence issues. GA given BW is fit with logistic regression with a cumulative logit link function.
|
