|
Activity Number:
|
252
- Replicate Weights and Variance Estimation
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2018 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #329023
|
Presentation
|
|
Title:
|
Strategies for Minimizing Unequal Weighting Effects in Two-Phase Sampling for Nonresponse
|
|
Author(s):
|
Dan Liao* and Paul Biemer and Darryl Cooney
|
|
Companies:
|
RTI International and RTI Internatinoal and RTI International
|
|
Keywords:
|
weight calibration;
regression trees;
mean square error;
paradata;
response propensity;
sample allocation
|
|
Abstract:
|
Many surveys today are employing two-phase sampling designs for nonresponse followup (NRFU) to save costs while increasing response rates and reducing nonresponse biases. Thus, rather than following up all nonrespondents when the survey reaches phase capacity, only a random subsample of the nonrespondents is followed up. One consequence of this strategy is weight variance inflation that may increase standard errors to the point where some of the nonresponse bias reduction benefits of NRFU are lost. This presentation discusses three weighting strategies that aim to maximize the reduction of nonresponse bias from two-phase NRFU while minimizing the weight variation to the extent possible. One strategy is based upon sample weighting the second phase sample to account for subsampling. The other two strategies are based upon the work of Singh, Iannacchione and Dever (2003) which incorporate composite weights that attempt to trade-off bias and variance. Results for the three strategies will be illustrated using the National Longitudinal Study of Adolescent to Adult Health (Add Health) Wave V sample which employed a mixed mode data collection protocol that culminated in a two-phase NRFU.
|
Authors who are presenting talks have a * after their name.
