Rare Binary Spatial Regression

• Activity Number: 687
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Samuel Morris* and Brian J. Reich
• Companies: North Carolina State University and North Carolina State University
• Keywords: Binary regresson ; Ozone ; Spatial dependence ; Rare events

Abstract

The selection of an appropriate link function for binary regression is important for accurate inference and prediction. Traditional link functions include logit, probit, and cloglog using independent models or Gaussian random effects to capture spatial dependence. These models allow for some amount of asymmetry in the data, but modeling spatial dependence using Gaussian random effects does not always capture extremal dependence. More recently, Wang and Dey (2010) developed the generalized extreme value (GEV) link function as a flexible framework for modeling independent binary data that are highly asymmetric. This paper builds on the GEV link function by incorporating spatial dependence via the hierarchical max-stable model for spatial extremes by Reich and Shaby (2012) to capture extremal dependence. We present a simulation study regarding the assumption of independence and link selection as well as an application to modeling ozone compliance at sites throughout the United States.