Conversation
dsheldon
left a comment
dsheldon
left a comment
There was a problem hiding this comment.
Nice work, @Serena-Wang! Most of the changes seem spot on. (I didn't check pandas manipulations very carefully.)
The only comments of significance were about the seeding process and the implications of dropping the first cumulative observation in SEIRD.py and SEIRD_renewal.py.
Seeding has always been a little fuzzy. My best guess is the changes are insignificant for incident forecasts but could matter for cumulative forecasts. I think we probably need to discuss exactly what we want there. It would be good if you could test what happens with cumulative forecasts, esp. with a start date well into the pandemic, to see what happens.
| death0 = death[0] | ||
| death = clean_daily_obs(onp.diff(death)) | ||
| death = clean_daily_obs(death[1:]) | ||
|
|
There was a problem hiding this comment.
I have a similar comment to the one in SEIRD.py --- how does this affect seeding and cumulative forecasts?
| @@ -1,3 +1,4 @@ | |||
| from pickle import FALSE | |||
elray1
left a comment
elray1
left a comment
There was a problem hiding this comment.
minimal comments other than what Dan already noted.
| else: | ||
| variables = ['y', 'z'] | ||
| observations= [confirmed, death] | ||
| observations= [confirmed.cumsum(), death.cumsum()] |
There was a problem hiding this comment.
Flagging this as a place where we might want to add in cum death on date before start_date
| util.redistribute(data['TX']['data'], '2021-03-26', 2029+294, 90, 'confirmed') | ||
|
|
||
| util.redistribute(data['MN']['data'], '2021-03-25', 20, 20, 'death') | ||
| #util.redistribute(data['MN']['data'], '2021-03-25', 20, 20, 'death') |
There was a problem hiding this comment.
did we just decide this was no longer necessary?
| # https://github.com/CSSEGISandData/COVID-19/issues/3705 | ||
| # (backdistributed week of March 1) | ||
| #util.redistribute(data['IA']['data'], '2021-02-19', 26775, 200, 'confirmed') | ||
| util.redistribute(data['IA']['data'], '2021-02-19', 26775, 200, 'confirmed') |
There was a problem hiding this comment.
this was previously commented out but isn't anymore?
…run_model and config, fix typo
3 participants
mechbayes/jhu.py:load_us_covidcast()to load data for all target variables from covidcast.scripts/run_model.py, add boolean parameteruse_hosp_as_deathto indicate whether a model needs to replace death observations with hospitalization observations