|
Activity Number:
|
389
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301772 |
|
Title:
|
Using the Peters-Belson Method to Measure Disparity in Complex Surveys
|
|
Author(s):
|
Sowmya R. Rao*+ and Barry I. Graubard and Joseph L. Gastwirth
|
|
Companies:
|
National Cancer Institute and National Cancer Institute and George Washington University
|
|
Address:
|
EPS 8047, 6120 Executive Blvd., MSC 7244, Bethesda, MD, 20892,
|
|
Keywords:
|
disparity ; survey methods ; Taylor linearization variance ; delete-one-group jackknife variance ; logistic regression ; multiple linear regression
|
|
Abstract:
|
The Peters-Belson (PB) approach used to assess wage discrimination can also be used to analyze disparities in large surveys for a variety of health and social behaviors including cancer-screening. This approach fits a regression model to the majority group and estimates the expected value a minority individual would have from this equation. The average difference between observed and expected values for the minority members estimates the disparity not explained by the covariates. Because of the complex sample designs of surveys, we show how to apply the Taylor linearization and delete-one-group jackknife methods to estimate standard errors for the disparity. The 1998 National Health interview Survey on cancer-screening among women is used for illustration. We found that the Taylor and jackknife methods gave similar standard errors. Although the jackknife method is easy to apply, it is more computer intensive than the Taylor method. The Taylor linearization method can be easily implemented using the existing software packages with the appropriate commands to account for the complex survey design.
|
