Abstract:
|
In response to challenges associated with traditional probability sampling, survey organizations have started using nonprobability samples to supplement probability samples in order to improve cost efficiency and timeliness of data dissemination. Nonprobability samples are cheaper but they may be biased. To correct for potential bias of nonprobability samples, researchers have proposed and implemented a range of methods to model the survey weights, the survey response variables, or both. Common methods include calibration, statistical matching, propensity weighting, small area estimation, and so on. No standard method has emerged so far. This roundtable offers an opportunity for interested researchers to share their most recent research and experience in combining probability and nonprobability samples for survey estimation. Discussions will focus on successes, challenges, client communications, and future research agenda.
|