search

jarad / FluSight

An R package containing functions used in the CDC Flu Forecasting competition
GNU General Public License v3.0
12 stars 7 forks source link

Use of ILI vs wILI in create_truth #23

Closed elray1 closed 6 years ago

elray1 commented 6 years ago

Documentation for create_truth says that "For ‘challege = "ilinet"’ or ‘challenge = "state_ili"’, must contain columns location, week, and wILI."

The verify_ILI_colnames function sets the valid names to the following (last element is "ILI" rather than "wILI"): valid_names <- c("location", "week", "ILI")

Additionally, the ILI column is used throughout the create_truth, create_onset, create_peak, and create_week functions.

Probably we really want a wILI column, or to allow the user to provide wILI and then map that to whatever gets used in the computations of onset/peak timing and incidence forecasts?

craigjmcgowan commented 6 years ago

When I added the functionality for the state-level truth, I changed it to ILI since CDC is not publishing weighted ILI measures for state level data. Not changing the wILI in when fluview = T and not fixing the documentation was an oversight on my part. I think I'd like to keep it more general as ILI in the "create" functions but I like the idea of having some code in there to convert wILI to ILI if the user supplies a dataset with wILI instead.

CDC has also updated the FluView portal to HTML5 which will allow for us to pull state ILI data as well. I put a pull request in to update the cdcfluview package and once that's accepted I'll update the create_truth code to take advantage of being able to pull that data in automatically.