In response to the need to estimate CV19 prevalence using sample survey data with properties that make traditional frequentist analysis difficult, we combined existing Bayesian approaches to build a new integrated Bayesian approach to solve these challenges:
- Very low response including sampling units with no responses at all,
- Very few positive cases,
- Varying number of results from multiple low-quality antibody tests for each participant,
- Antibody tests whose performance characteristics were very poorly described, and
- The possibility of selective response.
The model is fully described in this preprint and article in Proceedings of the National Academy of Sciences.
- David Kline david.kline@osumc.edu
- Richard Li lizehang@ucsc.edu
- Yue Chue chu.282@buckeyemail.osu.edu
- Jon Wakefield jonno@uw.edu
- Abigail Norris Turner ant@osumc.edu
- Samuel Clark work@samclark.net