Allow the user to pass in weekly count data

CDCgov / ww-inference-model

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

https://cdcgov.github.io/ww-inference-model/

Apache License 2.0

12 stars 1 forks source link

Allow the user to pass in weekly count data #93

Open kaitejohnson opened 3 weeks ago

kaitejohnson commented 3 weeks ago

Goal

Current formulation requires count data to be daily. Want to allow the user to pass in weekly count data.

Requirements

[ ] allow the user to specify that the counts being passed in are weekly sums (not just observations on a single day of the week)
[ ] modify model to either take in daily or weekly data, and produce either daily or weekly forecasts

Context

@akeyel 's suggestion + likely necessary change for shifting reporting requirements

seabbs commented 2 weeks ago

I'd suggest you do this by adding an additional accumulate functionality vs changing the input (i.e daily but with NAs and in model aggregation). You can then add helper functions for users to make specifying this easier.

kaitejohnson commented 2 weeks ago

So suggestion would be if the user has weekly data, provide helper functions and examples in the vignette to show how to convert weekly data to daily to pass into the wwinference() wrapper function.

I would think the best thing to do here would be to set the weekday effect rather than estimate it? So we would want the user to indicate that the data in its original form was weekly rather than just say a very weird reporting effect of daily data.

seabbs commented 2 weeks ago

you can do it that way but as its multiplicative it won't quite be right. Its better to explicitly aggregate