|
Activity Number:
|
41
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 3, 2003 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301172 |
|
Title:
|
Two Weighting Schemes for Combining Sample in the Health Behaviors in School-age Children Survey
|
|
Author(s):
|
William H. Robb*+ and Ronaldo Iachan and Pedro J. Saavedra
|
|
Companies:
|
Macro International, Inc. and ORC Macro International and ORC Marco
|
|
Address:
|
126 College St., Burlington, VT, 05401-8456,
|
|
Keywords:
|
sampling ; weighting ; variance ; schools ; HBSC
|
|
Abstract:
|
The Health Behaviors in School-age Children (HBSC) is a national survey of sixth to tenth grade students which is part of an international study. The HBSC requires confidence intervals for the population as a whole, as well as for minority populations. In previous cycles, oversampling minorities was accomplished with one sample that increased sampling probabilities for schools and PSUS with large minority enrollments. For the latest cycle a cluster sample was designed to be self-weighting, and a supplement sample was designed to over-sample minorities. Both samples selected PSUs with Probabilities Proportional to Size; for the supplement sample, the size measure assigned higher probabilities to schools with greater concentrations of minorities. This paper compares two schemes used to weight the combined sample. One scheme calculated weights separately, and created a linear combination with coefficients inversely proportional to the effective sample size. The other scheme calculated an overall probability of inclusion in one or both samples. Variance estimates were compared to examine the differences between weighting schemes.
|
