Hello, I have a question regarding the volume anomalies test. I created a volume test in a table without a timestamp column. For this test, I set a training period of 3 days and a detection period of 1 day, as specified in the .yml configuration file.

data_tests:
      - elementary.volume_anomalies:
          training_period:
            period: day
            count: 3
          detection_period:
            period: day
            count: 1

this is the

metric_properties

from

elementary.data_monitoring_metrics

{
  "dimensions": null,
  "event_timestamp_column": null,
  "freshness_column": null,
  "time_bucket": {
    "count": 1,
    "period": "day"
  },
  "timestamp_column": null,
  "where_expression": null
}

Upon examining the

elementary.metrics_anomaly_score

table, I noticed that the

training_avg

and

training_stddev

values are higher than expected. It appears that the calculation is not considering the specified 3-day training period from

elementary.data_monitoring_metrics

. As you can see the count of the table this month is 35 rows (used to have ~1000 rows in the previous months). Today I reduced it to 10 rows and later to 1 row to see if the test will fail but as expected it doesn't fail as the

training_avg

is too big (I was expected to be 35). I was wondering if I miss anything else from the configuration to correctly calculate the

training_avg

and

training_stddev

so the anomlay score can be calculated correctly. Thank you!