An identical _has_outliers function exists in:
engine/missing.pyengine/model_select.py
Both call detection_bounds(s, "iqr", 1.5) with a hardcoded factor. If one is updated and the other isn't, the imputation ranking (rank_missing_models) and the actual imputation logic (_handle_low) will disagree on whether a column has outliers — leading to the report saying "mean is appropriate for this normal distribution" while the actual fill uses "median".
Fix: Consolidate into a single function in a shared location (e.g. steps/outliers.py) and import it in both modules.
An identical
_has_outliersfunction exists in:engine/missing.pyengine/model_select.pyBoth call
detection_bounds(s, "iqr", 1.5)with a hardcoded factor. If one is updated and the other isn't, the imputation ranking (rank_missing_models) and the actual imputation logic (_handle_low) will disagree on whether a column has outliers — leading to the report saying "mean is appropriate for this normal distribution" while the actual fill uses "median".Fix: Consolidate into a single function in a shared location (e.g.
steps/outliers.py) and import it in both modules.