Speaker
Description
Unlike other areas of modern statistical inference, genomic epidemiology lacks theory to guide decisions about how to sample pathogen genomes. This makes it difficult to answer even simple questions about how sampling choices impact the quality of inferences drawn from genomic data. Researchers therefore often resort to sampling opportunistically, leading to inefficient and unrepresentative sampling which can further bias downstream inferences. We therefore consider the problem of designing optimal sampling strategies that maximize the information gained from genomic data while minimizing total sampling effort. By formulating the optimal sampling problem as a Markov decision process (MDP), we can jointly model the epidemic dynamics of a pathogen together with sampling within a sequential decision making framework. Combining MDPs with dynamic programming allows optimal sampling strategies to be learned in several common epidemiological scenarios, including estimating epidemic growth rates and transmission rates within and between different host populations.
| Expedited Notification | No thanks, I do not require Expedited Notification |
|---|
