|
Activity Number:
|
292
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #300057 |
|
Title:
|
NMAR Nonresponse with Limited Covariate Information
|
|
Author(s):
|
Danny Pfeffermann*+ and Anna Sikov and Michael Sverchkov
|
|
Companies:
|
Hebrew University & University of Southampon and The Hebrew University of Jerusalem and BAE Systems/Bureau of Labor Statistics
|
|
Address:
|
Department of Statistics, Jerusalem, 91095, Israel
|
|
Keywords:
|
Calibration ; Sample-distribution ; Sample-complement distribution ; NMAR ; H-T estimator
|
|
Abstract:
|
In this talk we utilize relationships between the population, the sample and the sample-complement distributions developed previously. These relationships allow expressing the distribution of the missing target outcomes as a function of the distribution under full response and the response process. The missing outcome distribution is used for imputing the missing values and for estimating the population mean of the outcomes. The proposed approach is extended to situations where the population means of the covariates are known, but the covariates themselves are unknown for the nonresponding units. In such situations the response process for the covariates cannot be ignored even if the outcome nonresponse given the covariates is missing at random (MAR), requiring a proper imputation of the missing covariates. We illustrate the proposed approach using simulations and a real data set.
|
