|
Title
|
Room
|
|
Model Misspecification, Missing Data and Measurement Error
|
M-Consulate
|
|
Date / Time
|
Sponsor
|
Type
|
|
08/05/2001
2:00 PM
-
3:50 PM
|
Biometrics Section*, Section on Survey Research Methods*, Section on Health Policy Statistics*, ENAR
|
Topic Contributed
|
|
Organizer:
|
Mari Palta, University of Wisconsin- Madison
|
|
Chair:
|
Sin-Ho Jung, Indiana University School of Medicine
|
|
Discussant:
|
|
|
Floor Discussion
|
3:45 PM
|
Description
Missing data, measurement error and unknown covariates are common in practice, especially in observational studies and in surveys. These practical problems have similar implications for model specification and can often be differerntiated only by subject matter knowledge. We specify some general structures to describe the problems, examine the implications of ignoring them, and then proceed to detection of the resulting misspecification and adjustment of estimators. A major focus is on longitudinal data, as they present special opportunities for detecting and correcting for model misspecification. We emphasize robust methods that are easy to implement in practice. We also discuss a specific problem that arises in the imputation of data in a contingency table.
|
|
