Mon, Jan 23, 2017 at noon:
H. Luke Shaefer
Groves, Robert M., James Wagner, and Emilia Peytcheva. 2007. "Use of Interviewer Judgments about Attributes of Selected Respondents in Post-Survey Adjustments for Unit Nonresponse: An Illustration with the National Survey of Family Growth." Joint Statistical Meetings Proceedings, Survey Research Methods,
Postsurvey adjustments based on propensity models are increasingly popular in survey sampling. After examining the how weighting classes based on discretized estimated propensities can act to reduce the mean square error of respondent-based estimates, it is straightforward to focus on the covariance of the estimated propensity, p, and the survey variables, y's. When propensities are related to the y variables, then adjustments are likely to alter the expected value of the adjusted estimates. Because of this fact, it is desirable to examine ways to collect on both respondents and nonrespondents information predictive of the y variables. We refer to these variables as z variables. This paper reports on one initial attempt to do so, in the context of the National Survey of Family Growth. It reviews the paradata structure of the National Survey of Family Growth and presents initial efforts to model such paradata in a useful way for postsurvey adjustment.