The Gains Chart of GeneXproTools is quantile-based and shows the
cumulative gain as more cases are included in a campaign or test. The
Lift Curve is compared to both a Random Model and an
Ideal Model, showing clearly the advantages of using a model as opposed to not using one.
The Random line in the Gains Chart represents the average response
rate. And the Ideal line represents a perfect model that is never
wrong and therefore could select all the estimated positive
responses. So, the further up (or down, for inverted models) the
Lift Curve is from the Random Line the better the model.
The Gains Ranking Quality (GRQ) is a good indicator of the quality of a model. It is defined as the relation between the area under the Ideal Model and the area under the
Lift Curve. It ranges from -1 to +1, with zero corresponding to the Random Model. The better the model the closer the GRQ gets to either +1 or -1 (for inverted perfect models GRQ = -1, whereas for normal perfect models GRQ = 1). As an additional quality measure, the
Area Under the
Lift Curve (represented by AUC in the Gains Chart) is also evaluated and shown both in the chart and the companion
Gains Statistics Report.