Fixing issues with inference and NAs

[saraloo]

Describe your changes.

• Fixed reading in ground truth data NAs as 0 -> removed this
• Changed filtering of dates so that we only compare the same simulation/observed dates (post aggregation) and only when observed values are not NA

What does your pull request address? Tag relevant issues.
#286 and related prior issue #272 (do need to filter dates to be the same but was just in the wrong spot)

Mentions of relevant team members.
@shauntruelove @alsnhll

[saraloo]

Running tests to ensure this is working

[saraloo]

Summary of changes:

• Remove initial adjustment of NAs in groundtruth to 0s
• Include filtering to only compare ground truth and simulation output with same date, and that are not NA
• Change aggregation date to the end of the week (Saturday) rather than first date of the week to better match with typical ground truth data

[saraloo]

Added an quick fix change to change the default of the total aggregation likelihood to FALSE

A later enhancement would be to add these sorts of options to the config - eg aggregation dates (in the emcee version, there is a python W-SAT option), and any summed likelihoods considered, or other modifications...

[jcblemai]

Very important bug fixes

[TimothyWillard]

We will need to come back and add unit tests to demonstrate that this change did in fact fix the underlying issue so don't close the corresponding issue yet.

[saraloo]

Yeap tests for inference are 100% required. would be great. Thanks both for review

…

On Fri, 13 Sept 2024, 10:39 am Timothy Willard, ***@***.***> wrote: ***@***.**** approved this pull request. We will need to come back and add unit tests to demonstrate that this change did in fact fix the underlying issue so don't close the corresponding issue yet. — Reply to this email directly, view it on GitHub <#292 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKZGJPXMT4ZWMVJUKCNCEGDZWMBM5AVCNFSM6AAAAABMG62472VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDGMBTGQ3DGMJSGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[jcblemai]

agree on the test @TimothyWillard, but thanks Sara for finding and fixing this