An in-development R package and a Bayesian hierarchical model jointly fitting multiple "local" wastewater data streams and "global" case count data to produce nowcasts and forecasts of both observations
Initially thought passing in natural scale would be easiest, but realizing with #97 and handling of the LOD, that we want to provide helper functions that show the user how to replace any censored data to a value below the LOD to pass into the model. Makes sense to then put this in log scale because you wont have errors if they put 0. Also makes sense because the default figures present the data in log scale
Requirements
[ ] refactor such that wwinference() and upstream functions expect log scale concentrations
[ ] edit checkers and validate accordingly
[ ] explicitly state in the vignette that the user must pass in values with no missingness for wastewater concentration, even for those below the LOD. Explicitly state they should not exclude these either, because the below LOD values still inform the model via the censoring implementation.
For a subsequent PR:
[ ] example using helper function to replace NA with a value that is below the LOD (with a default of replacing it with half the LOD, as NWSS does) (#97 )
Goal
Initially thought passing in natural scale would be easiest, but realizing with #97 and handling of the LOD, that we want to provide helper functions that show the user how to replace any censored data to a value below the LOD to pass into the model. Makes sense to then put this in log scale because you wont have errors if they put 0. Also makes sense because the default figures present the data in log scale
Requirements
wwinference()
and upstream functions expect log scale concentrationscheckers
andvalidate
accordinglyFor a subsequent PR: