Ellipsoids create an ellipsoid that shows the region within which, with a defined confidence percentage, the software expects points to appear in relation to the mapped positional dimensions. If a categorical feature has been mapped to color, an ellipsoid will be generated for each color. If instead, a numerical feature is mapped, then an ellipsoid will be created for each color bin.
Required Features and Dimensions
Ellipsoid requires at least two numerical features mapped to two of the positional dimensions (for 2D ellipses) or three dimensions (for 3D ellipsoids).
Mapping only to the positional dimensions will create an ellipsoid for the entire visualization. To visualize multiple distributions, map a numerical or categorical feature to the Color dimension. Each category in that feature will have its own ellipsoid.
Ellipsoid settings, found on the Input tab of the Mapping panel, allow you to customize your visualization:
- Confidence Percentage – This dropdown allows you to change the confidence percentage of the ellipsoid. It defaults to 95%, indicating that the bounds of this ellipsoid are showing where, given the distribution, we would expect 95% of points to be located. Increasing the confidence percentage makes the ellipsoids bigger, while decreasing it makes them smaller. Note: a confidence distribution is not a predictive probability distribution, but seeing how a category interacts with other features is useful for inferring probability.
- Show/Hide Points – This dropdown allows you to hide points. This is useful for viewing just the ellipsoids in a crowded visualization.