Sensitivity (True Positive) Vs. 1- Specificity (True Negative) in Marketing Models

Sensitivity: The likelihood of heart attack, given that the patient's test is positive (posterior probability) - TRUE POSITIVE

Specificity: The likelihood of heart attack, given that the patient's test is negative; FALSE NEGATIVE

(1-specificity) is the likelihood of no heart attack given that the test is negative (again posterior probability) - TRUE NEGATIVE

Let us say "AGE" is the covariate. That is, we have a posterior parametric model with AGE as the independent variable.

A test based on AGE dominates another test (means better than the other), if the TRUE POSITIVES VS. TRUE NEGATIVES (discrimination curve - Receiver Operating Curve - ROC) is above the already established discrimination curve.

In traditional marketing context, we use sample deciles and TRUE positives as the lift curve. So we are likely to build great models which may have bad rates of "FALSE POSITIVES" and "FALSE NEGATIVES".

A great introduction about True positivies, True Negatives, False Positives, and False Negatives is http://www.cs.cornell.edu/courses/cs678/2006sp/performance_measures.4up.pdf