Model Selection for Categorical Data When the Interaction Effect Is Presented

• Activity Number: 568
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Pannapa Changpetch*+ and Dennis Lin
• Companies: Penn State University and Penn State University
• Keywords: Association rules analysis ; Interaction effects ; Logistic regression models ; Model selection

Abstract

Interaction effect is very common in reality but has received little attention in literature of categorical model. This is especially true for higher-order interactions. Conventionally, interactions are typically ignored. To overcome this problem, we propose a model selection procedure by implementing an association rules analysis. We do this by (1) exploring the combinations of input variables which have significant impacts to response via association rules analysis; (2) selecting the potential (low- and high-order) interactions; (3) converting these potential interactions into new dummy variables; and (4) performing variable selections among all the input variables and the newly created dummy variables (interactions) to build up the optimal logistic regression model. Our model selection procedure establishes the optimal combination of main effects and potential interactions. The comparisons are made through thorough simulations. It is shown that the proposed method outperforms existing methods in all cases.