|
Activity Number:
|
35
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #306823 |
|
Title:
|
Approaches to Obtaining Standard Errors for Parameter Estimates in Latent Class Analysis
|
|
Author(s):
|
David M. Thompson*+
|
|
Companies:
|
The University of Oklahoma
|
|
Address:
|
825 NW 49th Street, Health Sciences Center, Oklahoma City, OK, 73118,
|
|
Keywords:
|
latent class analysis ; standard errors ; E-M algorithm ; SAS-IML ; PROC Catmod
|
|
Abstract:
|
Latent class analysis (LCA) has attracted the interest of clinical professionals who must place clients in diagnostic or prognostic categories when a gold standard for doing so is poorly defined. However, the classic MLE approach to LCA employs an expectation-maximization algorithm that does not yield standard errors. SAS-IML or SAS PROC CATMOD's loglinear modeling facility permit LCA approaches that open up strategies for obtaining standard errors. These include repetitive analyses using a grid of initial estimates, or conversion of CATMOD's loglinear expressions for SE into probabilities. The presentation addresses data in which four binary indicators permit estimation of a two-class latent structure. CATMOD's flexibility in loglinear modeling potentially allows estimation of larger models, including ones that accommodate residual local dependence among indicators.
|
