Arize AI

Statistic Widgets

How to use a statistic widget on a dashboard


The statistic widget represents an aggregate value for a time period. The value shown represents what you would get if you summed up all the timeseries data points.

Use Cases

For analyzing aggregate statistics, such as accuracy, over a time period. The chart type selection determines the type of metrics selectable in the setup:
  • Model Data Metrics
    • Count, Average, or Percent / Percent Error
  • Evaluation Metrics
    • Accuracy, Precision, Recall, F1, Sensitivity, Specificity, False Positive Rate, False Negative Rate, MAE, MSE, RMSE, or MAPE
Data Metric vs Evaluation Metric

Data Metrics

The same options available on the time series are available for the statistics widget:
Statistic widget total for time period (left) vs Timeseries (right)
The predictions count below shows the configuration and aggregation options for a statistic widget:
Configuration for Statistic Widget

Evaluation Metrics

The same evaluation metrics available in the timeseries plots are available in statistic widgets. Here's an example showing a statistic widget on the left representing the accuracy for the entire time period.
Statistic Widget Total for time period (right) vs Timeseries By Day (left)
The statistic widget will have different configuration options based on the selection of an evaluation metric.
Statistic Widget Confgiuration


When you select Validation or Training the data displayed is based on the date of the batch not the production date range selector. This allows teams to compare a batch of data from training to any time of production.
Environment selection
In the picture above the fixed date the data was sent is shown below the environment selection.
Questions? Email us at [email protected] or Slack us in the #arize-support channel