|
Activity Number:
|
153
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #309052 |
|
Title:
|
A New Algorithm for High-Dimensional Calibration in Observational Studies
|
|
Author(s):
|
Michael E. Jones*+ and Ismael Flores Cervantes and David R. Judkins
|
|
Companies:
|
Westat and Westat and Westat
|
|
Address:
|
1650 Research Blvd, Rockville, MD, 20850,
|
|
Keywords:
|
poststratification ; raking ; weighting
|
|
Abstract:
|
It has been shown recently that raking can be used as a balancing technique when applying propensity scoring to reduce bias in quasi experiments. When raking to balance, the potential number of control variables may increase to a level where time to convergence is impractical, even with high powered computers. As an alternative to the classic raking procedure, we describe a new method designed specifically for high-dimensional calibration. We evaluate the new method primarily in terms of convergence speed. However, we are mindful that this tool could facilitate questionable estimation strategies wherein people calibrate on many extraneous dimensions. To assess the consequences of such a strategy, we evaluate the performance of an estimator of a survey mean based on weights that were calibrated on many dimensions that are independent of the variable of interest.
|
