Explain: a tool for exploring Maxent models

Written by Steven Phillips, with support from AT&T Labs-Research.

The explain tool shows how an additive Maxent model (i.e., one made without product features) is composed of separate contributions from each predictor variable.

Click on the map to get data for a single point. The corresponding point is shown on the response curve for each predictor variable on the right.

The plot labeled "Contribution to logit" shows the breakdown of the logit of the Maxent prediction, giving the contribution of each variable. The variables that are most limiting (having most negative logit) or most supportive of presence of the species (having most positive logit) can be readily seen from in this plot.

To zoom in on the map, sweep out an area with the mouse. The map can also be shifted or zoomed using the navigation buttons at the top left.