Abstract:
|
Predicting survey costs is of great interest to survey managers for efficiency gains in data collection. In the responsive survey design context, case-level predictions of costs can be used as inputs to case-level decisions. Improved cost predictions may enhance the quality of decisions that use these predictions as inputs. However, case-level costs are less likely to be recorded directly by interviewers; instead, interviewers report their total hours worked. Effort indicators are proxy indicators of survey costs. For example, the number of call attempts to finalize a case is thought to be associated with total interviewer hours and total costs. We use data from the Health and Retirement Study to evaluate alternative prediction approaches in predicting call sequence length. The prediction approaches include prediction algorithms and traditional regression methods.
|