Skip to main content
How does graphing work?

Specifics into how the graphs use interpolation, downsampling, and grouped statistics

Masaki Yamada avatar
Written by Masaki Yamada
Updated over a year ago

Time-series Charts

Time normalization:

  • All time based charts are normalized to the Run Start time which calculates the Elapsed Run Time (ERT).

  • If there is data recorded prior the Run Start time, click View -> Show Pre-run start data to reveal the data prior to t=0.

Zoom:

  • For Y-axis zooming, click on the two axis numbers to set the zoom bounds.

  • For X-axis zooming, click and drag within the graph.

Downsampling and Interpolation:

  • In order to maintain a snappy interface, we downsample the vast amount of available raw data for the line charts.

  • Data is interpolated between time points to enable the calculations in formulas and grouped time-series statistics for runs that do not have identical data frequencies.

  • Higher resolution raw data will render as you increase the zoom and is also available through the data export.

Chart Splits:

  • In the View options you can select to split the data:

    • All-in-one

      • This will show all metrics and runs on a singular graph.

      • Each metric has a unique texture and each run has a unique color.

    • Split metrics

      • This creates a separate chart for each metric.

      • Each run has a unique color which is consistent across each chart.

    • Split Runs

      • This creates a separate chart for each run or run group.

      • Each metric has a unique color which is consistent across each chart.

    • Separate

      • Each metric and run is separated onto their own chart.

Grouping:

  • Use the 'Group by' option to categorize runs by metadata and access Invert's inter-run statistical comparison tools.

  • Line chart groups aggregate the run's interpolated data and display a solid line for the 50th percentile (median) values with a shaded range of the 16th to 84th percentiles. This ensures a distribution-agnostic analysis. Unlike mean and standard deviation, these percentiles provide a summary of both central tendency and variability within data. If the data is normally distributed these percentiles will represent 1 standard deviation.


Scatterplot Charts

  • Toggle to scatter charts under the View dropdown menu.

  • The x-axis is set to the run name by default. However, you can select any categorical or time-series aggregated metric available to the selected runs by clicking on the tile next to the X.

  • The y-axis metrics can be any numeric values. This includes time-series aggregations, single-point values (numeric metadata), and calculated metrics.

    • Time-series aggregation options include mean, minimum, maximum, standard deviation, first, last, and count.

Did this answer your question?