EvalC3 Online: How to use the Compare Models screen

What you can see

At the top

The color coded cases on the left (will) represent all those cases where the outcome was present/absent as expected.

The columns to the right represent each model that is being compared.

The color coded cells in these columns indicate which cases where that column model correctly identfieid the outcome.

The right column shows how many models did so altogether.

At the bottom

At the bottom of the table the measures there describe the overall performance of each model:

• Number of True Positives = the number of cases where the model correctly identified that the outcome was as expected

• % Raw Coverage = the percentage of all cases with the expected outcome (i.e. TP+FN) that each model identified

• % Unique Coverage = the percentage of all cases with the expected outcome (i.e. TP+FN) that each model identified, but which no other model identified.

• % Aggregate Coverage = the percentage of all cases with the expected outcome (i.e. TP+FN) that all the models together identified

What you can do

Use the "# of models" column on the right to find cases predicted by more than one model. Within-case investigations of these models could help to find out which model has the most explanatory power, and to adjust the assessments of the comparative coverage of the different models