Creating Gains Tables and Lift Charts Using R

• Activity Number: 368
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Craig Rolling*+
• Companies: University of Minnesota
• Keywords: model selection ; R ; gains ; lift ; visualization ; marketing

Abstract

In marketing and finance, gains tables (also known as lift tables or lift charts) are commonly used to evaluate the performance of a statistical model or prediction algorithm. Gains tables rank observations by their predicted response values, create subgroups (typically deciles) of observations with similar rankings, and display the average response for each group. Gains tables show the accuracy of a model's predictions within different subgroups and indicate how well the model is able to discriminate high-response groups from low-response groups. They are especially useful for applications in which a prioritization of individuals (for example, to receive a treatment) is needed. This talk and poster will demonstrate a new easy-to-use R package that creates gains tables and their graphical equivalents.