|
Activity Number:
|
36
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #307469 |
|
Title:
|
An Application of Propensity Modeling To Adjust Weights for Nonresponse: Effectiveness of Restricting Variables and Propensity Values
|
|
Author(s):
|
Frank Potter*+ and Nuria Diaz-Tena and Stephen R. Williams
|
|
Companies:
|
Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc. and Mathematica Policy Research, Inc.
|
|
Address:
|
600 Alexander Park, Princeton, NJ, 08543,
|
|
Keywords:
|
non-response ; weighting ; propensity modeling ; survey bias and precision ; community tracking study
|
|
Abstract:
|
Adjusting for nonresponse can use logistic regression models to predict the probability that a unit will respond. The predicted propensities can reflect more predictive variables than in the weighting class method and the inverse of the propensity can be the weight adjustment factor. Having used this method for rounds two and three of a large physician survey, this paper describes the results from round four. The independent variables used in round four are expanded to include design variables, basic sampling weights, and higher-order interactions. Predictive power of the propensity models were substantially improved but also presented some interesting issues. The more effective models produce more extreme adjustment factors. We consider whether or not to restrict the nonresponse ranges of candidate variables and whether the extreme adjustments or the final weights should be trimmed.
|
