@kapa.ai Can you write a sample anomaly test based on 2 dimensions ( country and date)

Thinking completed Here is a sample

dimension_anomalies

test using two dimensions —

country

and `date`:

models:
  - name: your_model_name
    config:
      elementary:
        timestamp_column: updated_at
    data_tests:
      - elementary.dimension_anomalies:
          arguments:
            dimensions:
              - country
              - date
          config:
            tags: ["elementary"]
            severity: warn

A few important things to keep in mind when using two dimensions together: [dimensions config] • Combinations are watched, not individual columns. Elementary joins the two values with

;

and monitors every combination it finds — e.g.,

US; 2026-01-01

,

DE; 2026-01-01

, etc. This is not the same as monitoring

country

and

date

separately. [understanding dimension anomalies] •

date

is likely high-cardinality. Since

date

can have many distinct values, combining it with

country

will produce a very large number of combinations (e.g., 20 countries × 365 dates = 7,300 combinations). This can lead to significant noise and false alarms. It is generally recommended to use low-cardinality fields as dimensions. [dimension_anomalies test] • Order matters. If you ever reorder, add, or remove a column from the

dimensions

list, Elementary will lose its historical data and start collecting from scratch. [understanding dimension anomalies] • To monitor each field independently, write two separate tests instead of listing both columns in one test. 👍 0 button 👎 0 button ℹ️ Tip: Mention @elementary.ai in the thread for follow-ups or DM the app for 1-to-1 help.