Hi there, I'm currently trying to setup up some volume anomaly detection. I have this as my test:

data_tests:
      - elementary.volume_anomalies:
          timestamp_column: event_timestamp
          seasonality: day_of_week
          fail_on_zero: true
          training_period:
            period: day
            count: 7
          anomaly_sensitivity: 0.5
          tags: elementary

But it keeps erroring with an average which is not representative of the actual average. I've looked into what is executing and it seems that it is setting the buckets to start earlier than expected. It is starting the bucket from here: timestamp_trunc(cast(cast('2025-04-08 000000' as timestamp) as timestamp), day) But that doesn't really make sense as the window is only supposed to be 7 days? You can see the error in the average that it is generating:

The last row_count value is 13492. The average for this metric is 1927.429.

Has anyone seen this setting of the window go wrong? It just seems like it's pulling a date out at random. Thanks for any help!