|
Activity Number:
|
21
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 8, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Health Policy Statistics
|
| Abstract - #302095 |
|
Title:
|
The Effects of Correcting for Sample Selection Bias in Internet Panel Surveys Based on Random Digit Dialing Sampling When Estimating the Demand for Preventative Health Care
|
|
Author(s):
|
Trudy A. Cameron and J. R. DeShazo*+ and J. Michael Dennis and Rick J. Lee
|
|
Companies:
|
University of Oregon and University of California, Los Angeles and Knowledge Networks, Inc. and Knowledge Networks, Inc.
|
|
Address:
|
3250 Public Policy Bldg. , Los Angeles, CA, 90095-1656,
|
|
Keywords:
|
sample selection bias ; nonresponse adjustment ; Knowledge Networks ; internet
|
|
Abstract:
|
When evaluating survey-based demand for preventative health care, this paper will (i) estimate the probabilities of self-selection and retention in the web-enabled research panel, and (ii) develop a statistical correction for differences observed in (i). The data sources are the telephone numbers sampled for the RDD recruitment for the Knowledge Networks research panel, Census long-form information; and sample disposition for cases sampled for a large-sample stated preference study. The authors model (i) presence in the final estimating sample for the stated preference survey, vs. (ii) absence from this sample as a consequence of nonrecruitment to the panel, attrition from the panel prior to targeting the survey, and nonresponse to the invitation to participate in the survey. We will identify systematic variation in "propensity to be present in the estimating sample." Preliminary specifications will generate a various fitted propensity measures (or inverse Mills ratio) to be used in an a theoretic fashion as additional regressors in the multinomial logit model that explains respondent demand for preventative health care.
|
