|
Activity Number:
|
64
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 29, 2007 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #308950 |
|
Title:
|
Inference for Intraclass Correlation Models with Missing Responses at Random
|
|
Author(s):
|
Kai Fun Yu*+ and Mixia Wu and Aiyi Liu
|
|
Companies:
|
National Institutes of Health and National Institutes of Health and National Institutes of Health
|
|
Address:
|
NICHD/DHH, Bldg 6100, Room 7B05, Bethesda, MD, 20892-7510,
|
|
Keywords:
|
intraclass correlation model ; contrast ; missing
|
|
Abstract:
|
Intraclass correlation models are popular choices for the treatment of the data from block design or cluster sampling and longitudinal data with an individual random effect. Since missing data often occur in practice, the maximum likelihood (ML) method becomes complicated and requires numerical iterative computations. Some simple exact tests and estimators are often required in some situations. In this paper, we consider an basic intraclass correlation models with missing data at random and propose a new method to construct exact test statistics and simultaneous confidence intervals for all linear contrast in means, which permits missing data with non-monotone pattern. A real example from the Calcium for Preeclampsia Prevention (CPEP) Study is provided for illustration.
|
