|
Activity Number:
|
479
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #309860 |
|
Title:
|
Adjusting for Missing Covariates Using the EM Algorithm for Logistic Regression Models: A Study of the Effect of Childhood Overweight on the Age of Onset of Menses
|
|
Author(s):
|
Yi Liu*+ and Elizabeth A. Stasny and Pamela Salsberry and Patricia Reagan
|
|
Companies:
|
The Ohio State University and The Ohio State University and The Ohio State University and The Ohio State University
|
|
Address:
|
484 Stinchcomb Dr APT 10, Columbus, OH, 43202,
|
|
Keywords:
|
Discrete time hazard model ; EM-algorithm
|
|
Abstract:
|
This paper investigated various approaches to correct for missing data in analyses that examine the influence of early weight states on age of onset of menses. A discrete time hazard model was used to analyze the dynamics of menses onset. The EM-algorithm was used to adjust for the missing values in one of the key covariates, the child's weight state at age 6/7 years. Three different models were constructed and compared by fitting different structures on the covariate with missing values and on the missing indicator. Compared to the model based on complete data, the model adjusted for missingness predicted that girls were about 2 percentage points, or 6 percent, more likely to experience early onset menses (before age 11 years). Thus the empirical predictions of the model are sensitive to the treatment of missing data on childhood overweight as a risk factor for the timing of menses.
|
