Model Selection with Mplot

• Activity Number: 662
• Type: Invited
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Samuel Mueller* and Garth Tarr and Alan H. Welsh
• Companies: University of Sydney and Australian National University and Australian National University
• Keywords: model selection ; graphical methods ; using R ; information criterion ; model selection curves ; fence

Abstract

This talk introduces the mplot R package, which provides a collection of functions to aid exploratory variable selection. The package contains fast routines to make available modified versions of the simplified adaptive fence procedure (Jiang et al., 2009, AOS) as well as other graphical tools such as variable inclusion plots and model selection curves (Mueller and Welsh, 2010, ISR; Murray et al, 2013, SIM). A browser based graphical user interface is provided to facilitate interaction with the results. These variable selection methods rely heavily on resampling. Fast performance for standard linear models is achieved using the branch and bound algorithm provided by the leaps package. The graphical model selection methods in mplot visualise popular model selection criteria that involve minimizing a penalized function of the data over a typically very large set of models. The penalty in the criterion function is controlled by a tuning parameter which determines the properties of the procedure. The implemented methods in mplot allow us to better explore the stability of model selection criteria through model selection curves and this is demonstrated through case studies.