search

hubverse-org / flusight_hub_archive

Hubversion of FluSight 1 (2015-2019)
MIT License
2 stars 1 forks source link

how to define origin_date from epi year and week #19

Closed lmullany closed 2 months ago

lmullany commented 2 months ago

Based on #10 , we are going to derive an origin_date based on EWXX. From the original instructions

image

we, see that EWXX represents the latest MMWR week of data used in the forecast.. Some options for deriving origin_date:

  1. Set to the date of the first day of the epiweek represented by EWXX
  2. Set to the date of the last day of the epiweek represented by EWXX (i.e option 1 plus 6 days)
  3. Set to the date of the first day of the week following EWXX (ie. EWXX + 1) (i.e. option 1 plus 7 days)
lmullany commented 2 months ago

I think the choice depends on how we want users to interpret "horizon" within the files. The horizon is for the 1 week ahead, 2 week ahead, 3 week ahead, and 4 week ahead ili percentage forecasts. Which of the above three choices is best aligned with a reasonable relationship between this origin_date and the horizon ?

@nickreich any thoughts on this?

nickreich commented 2 months ago

The convention at the US COVID-19 Forecast Hub is to use the final Saturday of an epiweek as the target_end_date, so I think that would suggest option 2 might be the most consistent with prior behavior.

lmullany commented 2 months ago

Closing via #22