Fitting Logistic Regression Model Under Informative Sampling

• Activity Number: 132
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Korean International Statistical Society
• Author(s): MoonJung Cho* and Michael Sverchkov
• Companies: Bureau of Labor Statistics and Bureau of Labor Statistics
• Keywords: conditional inference ; generalized linear models ; non-response bias analysis ; US International Price Program ; weight smoothing

Abstract

We consider application of the relationship between the sample model and population model using method developed by Pfeffermann and Sverchkov (1999). Their results illustrated the potentially better performance of tests of distribution functions derived from their methods compared to the use of standard inverse probability weighting. We apply their methods to the non-response bias analysis using data from the US International Price Program. In this setting, models which generate the population values belong to the family of generalized linear models. We estimate population moments from the corresponding sample moments and use an estimation procedure that does not require a full specification of the underlying distribution. We provide an application algorithm and then discuss challenges and strategies.