Delphi_utils geomapper: update to fips 2020 population estimates

[dshemetov]

Description

See #1319.

Changelog

• Update source FIPS population source file.
• Update the dependent crosswalks.
• Add a sanity check notebook and remove an outdated one.

Fixes

• Closes Update delphi_utils.geomapper to 2020 FIPS population data #1319
• Partially addresses JHU CSSE COVID-19 case/death data for Chugach County and Copper River County in Alaska #1320

[krivard]

Looks like there are a whole pile of indicator tests that depend on the previous population mappings 😖

Those will need to be updated before we can merge.

[chinandrew]

Looks like there are a whole pile of indicator tests that depend on the previous population mappings confounded

Those will need to be updated before we can merge.

If there's a use case to applying 2020 data to pre Apr 1 (e.g. if someone who isn't doing an as_of analysis believes the 2020 data is more accurate than the 2019 estimates for the true population on January 1, 2021, they're free to use that), we could possibly kill two birds with one stone by changing geomapper so it can be init with a census_year argument where people can specify 2020 or 2019. Then all the broken tests can be set to use 2019 since it's just test data.

[dshemetov]

That's an interesting idea, but the tests are written with hard coded population values and functions that don't take a geomapper argument :/

[dshemetov]

Do you think we should then keep two directories of crosswalk files? Seems like a nice conservative approach, in case the update breaks something for users.

[dshemetov]

Gonna go ahead and just fix the tests for now and open issues for geomapper extensions to support 2019 and 2020 population estimates. We can think about coming up with a cutoff date for when we use the 2019 values vs the 2020 values for prop then.

[dshemetov]

Alright, looks like we squashed all the hardcoded prop tests.

[nmdefries]

Looks good to me!