Outliers — statistical detectors | ATX AnalyticsTable Docs

Outliers — statistical detectorsNumeric, multi-column, categorical and date detectors, reading results, and writing the report onto a sheet. Statistical outlier detection over numeric and categorical columns: z-score, modified z-score (MAD), the IQR fence rule, isolation forest and DBSCAN for multi-dimensional points, plus rare-category, unexpected-category and category-drift checks, impossible dates, time gaps and day-of-week seasonality. Results are ranked by score with the column types it inferred, so the reason a row was flagged is visible.

Where to find it

Transform → Quality → Outliers

What this page covers

Anomalies and outliers — two different questions
Outliers — the three numeric detectors, and when each is the right one
Outliers — the two multi-column detectors
Outliers — rare values, unexpected values and drift
Outliers — impossible dates, missing loads and wrong-day batches
Outliers — reading the results without fooling yourself
Outliers — writing the report onto a sheet

Settings and controls documented

Preview pin (live, paged query result)
Static pin (full query result loaded)
Scanned data + outlier flags
Write the report to

Full page

The full page — procedures and the explanation of every setting — is available to signed-in users.

ATX Institute on LinkedIn

This site needs JavaScript for the full experience. Every page above is reachable without it.