|
Activity Number:
|
250
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306883 |
|
Title:
|
Predicting Exposure at a Specified Time Based on an Extended Randomized Regression Model for Interval-Censored Data
|
|
Author(s):
|
Robert Lyles*+ and Amita K. Manatunga and Renee Moore and Michele Marcus
|
|
Companies:
|
Emory University and Emory University and Emory University and Emory University
|
|
Address:
|
Department of Biostatistics, Atlanta, GA, 30322,
|
|
Keywords:
|
coarse data ; environmental epidemiology ; random effects ; reproductive health
|
|
Abstract:
|
Motivated by a study aiming to link decaying maternal serum levels of an environmental exposure with reproductive outcomes among offspring, we use a random regression model to predict unmeasured in-utero exposures. The model is complicated due to coarse exposure data, most of which is reported to the nearest integer. We treat these data as interval-censored realizations of underlying lognormal exposure levels and maximize the resulting integrated likelihood. We derive empirical Bayes and constrained Bayes predictions of in-utero exposures accounting for the censoring and use simulations to compare their performances with those of estimation approaches proposed in the reproductive health literature. In addition to simulation results, we provide an analysis of data consisting of longitudinal serum measurements of polybrominated biphenyls (PBBs) from the Michigan Female Health Study.
|
