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

How to determine surveillance week #7

Open jarad opened 7 years ago

jarad commented 7 years ago

How can we determine the surveillance week from a Date? I'm trying to write a function to write an entry file that automatically determines the week from the Date. In R, you can extract a week from a Date using format(date, "%U") but this doesn't seem to coincide with the Flu Challenge Guidelines which have 2016-11-07 using week 43 data, but format(date, "%U") gives week 45. If it was 44, that would be understandable as the current week uses the past week's data. But why is it two weeks off?

craigjmcgowan commented 7 years ago

A submission is based on the last week of available ILINet data. For submissions on 2016-11-07, the latest ILINet data will be published on 11-4 and will consist of surveillance from MMWR week 43. Teams get the weekend and Monday to run their models incorporating the newest data.

Subtracting 2 from the week extracted using format(date, "%U") should work, with corrections for weeks around the new year.

jarad commented 7 years ago

Okay. I'm using MMWRweek (my package) to find the MMWRweek. Then it looks like we need to subtract 9 days from the current date in order to have any submission from 2016-11-07 or the past week be EW43. Can you please verify what will happen around the new year? I believe submissions on submissions on Jan 09 will be EW52 and submissions on Jan 16 will be EW01.

craigjmcgowan commented 7 years ago

That's correct - January 9 will be based on EW52 and January 16 on EW01.

The ILINet updates are delayed a little around Thanksgiving and Christmas and I know in past years we have delayed submissions accordingly, usually to Wednesday instead of Monday. I will check with the surveillance team here to find out the details and will let you know.

jarad commented 7 years ago

Okay, I don't know if I want this function to deal with that situation. Currently the function gives a warning saying that it is an experimental feature. I could update it to only give the warning around Thanksgiving and Christmas.