|
Abstract:
|
In response to the challenges of lower response rates and higher costs, survey methodologists have developed responsive survey design (RSD). RSD identifies and monitors a set of indicators of cost and data quality, enabling informed decisions about design changes during data collection. However, most current implementations of RSD are ad hoc and fail to leverage prior knowledge of data collection with incoming real-time information. We attempt to address this gap by developing a rigorously applied decision rule for a single RSD decision. In this study, we propose a Bayesian model-based decision rule to limit effort on remaining cases in a data collection that are expensive and similar to cases that have been already interviewed. To test the proposed rule, we used existing data from the Health and Retirement Study as inputs to a simulation study. Cost and propensity models were estimated in a Bayesian fashion, eliciting priors from previous wave data. The simulation study suggests that the rule is able to identify expensive cases that have a low risk of compromising the accuracy of estimates of descriptive parameters if they are dropped in the early stages of data collection.
|
