|
Activity Number:
|
388
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #305831 |
|
Title:
|
A Latent Model for Highly Skewed and Grouped Data
|
|
Author(s):
|
Huichao Chen*+ and Amita K. Manatunga and Robert Lyles and Michele Marcus
|
|
Companies:
|
Emory University and Emory University and Emory University and Emory University
|
|
Address:
|
Department of Biostatistics, Atlanta, GA, 30322,
|
|
Keywords:
|
grouped data ; skewed ; one-sided likelihood ratio test ; Michigan female health study
|
|
Abstract:
|
Data arising from reproductive epidemiological studies often present analytic challenges. To model the non-smooth distribution of exposure consisting of heaps and coarse values, we propose a general latent model for highly skewed and grouped data. We assume that the observed exposure is determined by the value of an underlying unobservable continuous response that follows a Weibull distribution. To accommodate correlations in repeated true latent responses, we introduce general random effects from the power variance function (PVF) family of distributions (Hougaard, 2000). The resulting marginal likelihood has a closed form. The performance of the proposed model is supported by simulation studies and its application is illustrated using repeated polybrominated biphenyl (PBB) exposure data on participants of the Michigan Female Health Study.
|
