Why MAD Outlier Clipping Can Differ Across Cross-Sections

Summary

This support exchange compares a median absolute deviation (MAD) clipping calculation with a platform's outlier-removal module. The example method computes a series median, finds the median absolute distance from that value, and clips observations to bounds set by a multiple of that distance. A user reports that its output differs from the platform module.

The reply explains that the platform groups observations by date and applies the operation to each date's cross-section. This grouping choice changes the observations used to compute the median and MAD, so results need not match a calculation applied to a differently grouped series. The exchange does not specify the module's scaling convention, multiplier defaults, or handling of missing values, so matching those details may also matter when reproducing results.

Key ideas

• MAD clipping bounds observations using the median and median absolute deviation of a series.
• The platform applies its outlier treatment separately within each date's cross-section.
• Different grouping changes the reference distribution and can produce different clipped values.
• The exchange does not detail scaling, defaults, or missing-value behavior.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.