Fix issue #12: Implement edge case with empty covariate set for MINT …

[RickardKarl]

…(#18)

• Fix issue Implement edge case with empty covariate set for MINT #12: Implement edge case with empty covariate set for MINT

This commit addresses GitHub issue #12 and provides a fix for the related bug in issue #7 where MINT would crash with confusing error messages.

Changes:

1. Enable MINT to handle empty covariate sets (covariate_vars=[])

   • Added proper handling in polynomial feature generation
   • Updated independence test to handle 1D parameter arrays
   • Ensured model fitting always returns 2D arrays

2. Improved error handling for insufficient samples

   • Added validation to catch empty coefficient lists early
   • Provides clear, actionable error message instead of cryptic "Need at least one array to concatenate" error

3. Comprehensive test suite for edge cases

   • Tests for empty covariates with different outcome/treatment types
   • Tests for polynomial features with empty covariates
   • Tests for insufficient samples error handling

All existing tests pass. The fix is backward compatible.

Fixes #12
Related to #7

• Update mint.py

Add len(coef_treatment_mech) == 0 to ValueError check

• Update test_mint_edge_cases.py

Remove unnecessary check of error test strings

• Delete ISSUE_12_FIX_SUMMARY.md

• Update mint.py

Fix consistent output shape of fit_model_jax

• Fix array shapes

---