|
Activity Number:
|
445
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
International Indian Statistical Association
|
| Abstract - #302783 |
|
Title:
|
On Analyzing Routes of Exposures Obtained from Questionnaires
|
|
Author(s):
|
Jaya Satagopan*+
|
|
Companies:
|
Memorial Sloan-Kettering Cancer Center
|
|
Address:
|
Department of Epi & Biostatistics, New York, NY, 10021,
|
|
Keywords:
|
residual diagnostics ; hat matrix ; unbiased estimation ; variance inflation ; penalized regression ; questionnaire
|
|
Abstract:
|
The use of questionnaires to assess exposure and behavior information is ubiquitous in epidemiology research. User-defined summaries or composite features are obtained as linear combinations of multiple questionnaire items to obtain insights into routes of exposures impacting the outcome. Their effects can be estimated by including the composite features as explanatory variables in a model, ignoring data about the individual questionnaire items. Such analyses may have limited scope since they assume that the individual items do not contain any information over and above that contained in the composite features, and that the user-defined summaries measure the composite features accurately. This talk will examine a penalized regression technique to obtain unbiased and efficient estimates of the composite features, and illustrate the method using data from the Study of Nevi in Children.
|
