Residual Plot

If the data set contains cross-validation values generated by the MOP, they can be displayed using this plot. Prediction (cross-validation) and fitting (CoD) values are displayed together, to illustrate the quality of prediction. This plot can be useful to identify and deactivate outliers and, as a result, improve the quality of prognosis of the surrogate models.

The residual plot has four different modes which can be selected in the plot settings:

  • Values: Shows the predicted and fitting values versus the original data values

  • Errors: Shows the predicted error and fitting error values versus the original values

  • Absolute Errors: Shows the absolute predicted error and absolute fitting error values versus the original values

  • Sample CoPs: Shows the contribution of each sample to the global CoP, the mean of all sample CoPs is the global CoP value

If the values are shown, the diagonal represents the perfect model which means that the predicted or fitted values are equal to the original data values. This corresponds to the horizontal line at zero when showing the error values. The error ranges (red lines) are scaled by the standard deviation (root mean square error) of the prediction error values (predicted minus original values) times the user-defined sigma level.

Statistical Data

This section of the plot displays additional information about statistic values (maximum, mean, rms) of the two different sets of error values, and can be compared. The CoP and the CoD of the currently selected model is also shown here.

Settings
OptionDescription
Common Settings
Show ordinate asSelect one of the following:

  • Values

  • Errors

  • Absolute errors

  • Sample CoPs

Sigma levelThe plot displays the mean line plus two sigma lines. This value defines the distance between them.
Preferences

The following preference settings are available:

  • Appearance

    • Font size

    • Line width

    • Symbol size

    • Enable axes

  • Axis Formats

  • Axis Ranges

  • Labels

  • Legend

  • Log Axes

  • Window appearance

For more details, see Plot Preference Settings.

Python scripting

Create Visual

Creates an Residual plot using data with data_id 

    residual_plot = Visuals.Residual(Id("Residual plot"), data_id  )

Add to Postprocessing

Adds Residual plot in postprocessing to control_container, using the specified relative positioning. 

    control_container.add_control (
        residual_plot,
        True,
        RELATIVE_POSITIONING,
        1/2., 0, 1/2., 1/2.
    )