# Leveraging Predictive Modelling from Multiple Sources of Big Data to Improve Sample Efficiency and Reduce Survey Nonresponse Error

Status: published
Visibility: public-pdf
Authors: with David Dutwin, Patrick Coyle, Ipek Bilgen, and Ned English
Venue: Journal of Survey Statistics and Methodology
Year: 2023
DOI: 10.1093/jssam/smad016
Canonical URL: https://doi.org/10.1093/jssam/smad016

## Summary

The question in this paper is whether big-data classifiers can improve sample targeting for hard-to-reach and low-incidence survey populations. Using 15 classifiers trained on a large probability-based panel and tested across three survey datasets, the analysis shows that the models generally outperform geographic clustering and often improve on vendor flags, usually by trading some coverage for noticeably higher incidence. Those gains vary across targets, making classifier-based sampling most useful when researchers care about efficiently finding specific subpopulations.

## Public Files

- Metadata: https://jyl19.github.io/papers/leveraging-predictive-modelling-from-multiple-sources-of-big-data-to-improve-sample-efficiency-and-reduce-survey-nonresponse-error/metadata.json
- PDF: https://jyl19.github.io/papers/leveraging-predictive-modelling-from-multiple-sources-of-big-data-to-improve-sample-efficiency-and-reduce-survey-nonresponse-error/paper.pdf