naobservatory/p2ra

Stats: Gaussian process model of prevalence timeseries

Closed this issue · 4 comments

dp-rice commented

In cases where our prevalence, MGS data, or both are timeseries, it would be natural to model the true prevalence as a gaussian process. It looks like Stan has good support for them, so we might as well try:
https://mc-stan.org/docs/stan-users-guide/gaussian-processes.html

dp-rice commented

This doesn't seem like a bottleneck right now, so wait until we need it.

dp-rice commented

This is a lot of work and does not seem to be a major problem with the model.