Activity Number:
|
241
- SPEED: Statistics in Social Sciences and Survey Research Part 1
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 9, 2022 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Survey Research Methods Section
|
Abstract #323359
|
|
Title:
|
A Statistical Model Predicting Final Yield During Data Collection
|
Author(s):
|
Rui Jiao and Daniel Guzman and Sabrina Zhang* and Andrea Piesse
|
Companies:
|
WESTAT and META and Westat and WESTAT
|
Keywords:
|
response propensity;
interim cases;
logistic regression ;
classification tree;
calibration
|
Abstract:
|
It is often useful to predict the final yield of a survey operation while it is still in the field. We consider the longitudinal setting in which new units are selected to refresh an existing sample. Survey paradata provide information about all contact attempts for sample units. Historical paradata capture a full picture of response behavior under a specified protocol, but they cannot fully predict final response for the current data collection because they do not account for temporal trends in survey response over time. Although the current paradata shed some light on the present trend, the information available may be partial. This presentation proposes an approach to utilize the paradata in the past and at present. It structures the historical paradata so that the grand mean of final response propensity can be separated from the effects of field efforts, and the cumulative effects at different times during data collection can also be accounted for. A logistic regression model is used for model training and prediction, and a classification tree algorithm is used for predictor selection. The intercept of the model is then updated using the current paradata for prediction.
|
Authors who are presenting talks have a * after their name.