Some Asymptotics for Geostatistical Model Selection

• Activity Number: 249
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Hsin-Cheng Huang*+
• Companies: Academia Sinica
• Keywords: Akaike information criterion ; Bayesian information criterion ; conditional information criterion ; fixed domain asymptotic ; increasing domain asymptotic ; variable selection

Abstract

Information criteria such as AIC and BIC are often used in model selection. However, their asymptotic behaviors in geostatistical model selection have not been well studied, particularly under fixed domain asymptotics based on increasing dense data observed in a fixed and bounded region. In fact, some spatial dependent parameters can't be consistently estimated by maximum likelihood under fixed domain asymptotics, thus making the problem difficult to handle. In this talk, I will first show some asymptotic properties of the generalized information criterion, including AIC and BIC. In addition, I will introduce a class of conditional information criteria extended from the conditional AIC, and provide some numerical and theoretical justification.