Using Residual Plots to Identify Model Misspecifications

• Activity Number: 179 - Contributed Poster Presentations: Section on Statistical Education
• Type: Contributed
• Date/Time: Monday, July 31, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Education
• Author(s): Emily Nystrom* and Julia L Sharp and William C Bridges and Patrick Gerard and Colin Gallagher
• Companies: Clemson University and Colorado State University and Clemson University and Clemson University and Clemson University
• Keywords: Residual Plots ; Model Misspecifications ; Correlation ; Interaction ; Survey

Abstract

Residual plots may be used to identify challenges with model assumptions and to distinguish types of model misspecification (e.g., missing a variable in the model). A tutorial on identifying whether a variable is missing based on residual plots was developed. We focused our attention on variables that were missing from the model that may be uncorrelated, correlated, uncorrelated but interacting, and correlated and interacting with another predictor in the model. We provided participants with a survey to distinguish types of misspecification by using residual plots before and after the tutorial. Results from baseline and follow-up surveys were compared to examine the effectiveness of the tutorial. Our results showed that the percentage of correct responses regarding most misspecification types were slightly higher following the tutorial.