|
Activity Number:
|
172
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #304257 |
|
Title:
|
Analysis of Dichotomized Responses in Longitudinal Studies with Missing Data
|
|
Author(s):
|
Kaifeng Lu*+
|
|
Companies:
|
Merck & Co., Inc.
|
|
Address:
|
217 Vasser Dr., Piscataway, NJ, 08854,
|
|
Keywords:
|
data augmentation ; GEE ; MAR ; MCMC ; multiple imputation ; repeated measures
|
|
Abstract:
|
Often a binary variable is generated by dichotomizing an underlying continuous variable according to a prespecified threshold value. One can impute the missing data on the continuous scale before dichotomizing them into responder statuses on the binary scale. Treatment comparison can then be based on the observed or otherwise imputed responder statuses. Simulation studies are carried out to compare different missing data approaches. We show that imputation approaches on the binary scale can be biased or less powerful than those on the continuous scale, incorporating distributional assumptions in imputing the missing data may/may not increase the efficiency of parameter estimates, and fixing the parameters at the maximum likelihood estimates when imputing missing data based on the normal distribution of predicted values does not necessarily compromise the coverage of confidence intervals.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
