Abstract:
|
There are surveys for which meeting the target sample size or yield, possibly also for subgroups, is very important. Survey statisticians need to predict the final number of completes among cases that have been released but do not yet have a final response status (interim cases). Some interim cases may have been in the field longer than others and the nature of interim statuses may vary. Methodologists have developed statistical models to estimate the likelihood of a case being finalized or completed on the next contact attempt. However, the problem of predicting how many interim cases will result in completed interviews over the course of data collection has received less attention because this requires combining predictions across all future contact attempts, where the prediction at each contact attempt is conditional on the previous contact attempts that have been made. This paper examines response propensity based on number of contact attempts, and proposes an approach using survival analysis to predict the proportion of current interim cases that will respond by the end of data collection. Finally, results of implementing the approach on synthetic data are presented.
|