Activity Number:
|
154
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Biometrics Section
|
Abstract - #310286 |
Title:
|
An Iterative Least Squares Process To Obtain Unbiased Variance Estimates at the Second Stage of a Two-Stage Model for Longitudinal Data
|
Author(s):
|
Mary Bartholomew*+ and Chris Gennings
|
Companies:
|
Food and Drug Administration and Virginia Commonwealth University
|
Address:
|
7500 Stanish Place, Rockville, MD, 20855,
|
Keywords:
|
bias ; Nelder-Mead algorithm ; uncertainty ; subject-specific ; population-averaged
|
Abstract:
|
Two-stage models for longitudinal data are intuitively appealing because the modeler obtains a good picture of subject-specific models and the parameters of the population-averaged model. However, it was widely publicized that variance estimates for the population parameters based on variation among estimated subject-specific parameters were biased due to failure to account for uncertainty in the subject-specific estimates. Likelihood methods that correctly estimate the variance in the population model have been developed but the population-averaged parameters are obtained directly and often the subject-specific models are neglected when this approach is used. An iterative procedure using the Nelder-Mead algorithm to cycle between subject-specific estimates and population parameter estimates removes bias and was applied to well known orange tree growth data with good results.
|