ELEX-2763: Estimandizer update

[dmnapolitano]

Description

The changes in this PR refine the changes made in PR #59. Hopefully the code and logic are much cleaner now. Thanks to @lennybronner for making most of these changes lol and the helpful feedback 😄 🙌🏻

Note that, as a result of these changes, a change will need to be made to the test bed; specifically, in modeltestbed/redo.py, include_results_estimand=True will need to be added to the constructor. Previously, I was trying to avoid having to modify the test bed, but that lead to some ugliness 🙈. But this change is extremely minimal and the rest of the code is much cleaner now, so IMO, it's worth it.

Jira Ticket

ELEX-2763

Test Steps

tox and you can also modify the test bed per the above (until the test bed itself is modified, of course). Thanks!

[dmnapolitano]

Hi @lennybronner , I'm stuck and need help with the integration tests ☹️

The one failing is this one:

elexmodel 2021-11-02_VA_G --estimands=dem --office_id=G --geographic_unit_type=precinct --percent_reporting 60 --historical --aggregates=county_classification

The issue is that baseline_dem is present but not results_dem, so the Estimandizer is trying (reasonably) to estimandize a results_dem and failing due to a lack of dem estimandizer. What do we want to do in this situation? Do we want to copy baseline_dem to results_dem? This is being called from MockLiveDataHandler 🤔

[lennybronner]

This looks v good! Though atm the integration tests are failing. Also when I run a historical election all subunits are unexpected.

[dmnapolitano]

This looks v good! Though atm the integration tests are failing. Also when I run a historical election all subunits are unexpected.

Thanks! Sorry about the failing integration tests. I have a question about that in a comment for you (above) because I'm confused about how we're expecting that test to work 🤔

In regards to the unexpected subunits, can you provide an example for that that I can look into? 🤔

[lennybronner]

This looks v good! Though atm the integration tests are failing. Also when I run a historical election all subunits are unexpected.

Thanks! Sorry about the failing integration tests. I have a question about that in a comment for you (above) because I'm confused about how we're expecting that test to work 🤔

In regards to the unexpected subunits, can you provide an example for that that I can look into? 🤔

oh oops, sorry I missed that. How was this test passing earlier? 🤔

Here is an example of all units being unexpected:

elexmodel 2020-11-03_USA_G --office_id=P --estimands=party_vote_share_dem --geographic_unit_type=county --pi_method=nonparametric --percent_reporting=30 --aggregates=postal_code --historical