|
Activity Number:
|
320
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #306352 |
|
Title:
|
Penalized Latent Class Regression: Incorporating Scientific Knowledge into Measurement Models
|
|
Author(s):
|
Jeannie-Marie Sheppard*+ and Karen Bandeen-Roche and Peter Zandi and William Eaton
|
|
Companies:
|
Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health and Johns Hopkins Bloomberg School of Public Health
|
|
Address:
|
550 N. Broadway, Baltimore, MD, 21287,
|
|
Keywords:
|
latent variable ; measurement ; validity ; penalization ; depression ; nosology
|
|
Abstract:
|
A primary challenge in psychiatric epidemiology is nosology. To establish an association between any risk factor and a disorder, one must define that disorder first. Building on work by Houseman and colleagues, we will use a penalty in the fitting of latent class regression (LCR) models to produce depression subtypes based on polymorphisms of the serotonin transporter (5-HTT) gene and individuals' patterns of symptoms. While the relationship between this functional polymorphism and depression has not been elucidated, serotonin has been implicated in affective disorders, and we therefore hypothesize genotype may refine depression classification usefully. We will discuss methods to fit these models and report on their application to the subtyping of depression. Ultimately, we aim to augment existing tools to incorporate scientific knowledge into measurement modeling.
|
