Fix MissingDataError in multi_mra when features contain NaN

[drussellmrichie]

Problem

run_multi_mra / _run_multi_mra crashes with:

statsmodels.tools.sm_exceptions.MissingDataError: exog contains inf or nans

when any feature in ind_vars has NaN values. This is common with real-world property data — fields like bldg_parking_spaces, bldg_has_garage, bldg_central_air, etc. are sparsely recorded by assessors and arrive as NaN after the astype(float) conversion on line 1940.

Root cause

_run_multi_mra one-hot encodes categoricals and calls astype(float), which preserves NaN from nullable Float64 columns as float64 NaN. There is no imputation step before sm.OLS(y_train, X_train).fit(), so statsmodels raises immediately.

Fix

Impute NaN with training-set column medians immediately before the global OLS fit, and propagate the same medians to X_test, X_sales, and X_univ so all splits are treated consistently:

col_medians = X_train.median()
X_train = X_train.fillna(col_medians)
ds_prepped.X_test  = ds_prepped.X_test.fillna(col_medians)
ds_prepped.X_sales = ds_prepped.X_sales.fillna(col_medians)
ds_prepped.X_univ  = ds_prepped.X_univ.fillna(col_medians)

This is the same median-imputation pattern already applied in utilities/stats.py (calc_elastic_net_regularization, calc_p_values_recursive_drop, calc_t_values_recursive_drop, calc_vif_recursive_drop) and accepted upstream in PRs #313, #316, #318.

Test plan

• Run multi_mra against a dataset with at least one feature containing NaN values — confirm it completes without MissingDataError
• Confirm predictions are produced for all rows (train, test, sales, universe)
• Confirm no change in behaviour when all features are fully populated

🤖 Generated with Claude Code

[github-actions]

Thank you for your contribution.
Please sign our CLA at the following link:
Click here to sign the CLA.

No action is required from you in this PR thread. Once you have signed the CLA externally, a maintainer will verify your signature and record it here on your behalf by commenting:

---

I affirm that this contributor has signed the CLA

---

Russell Richie seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}