|
Activity Number:
|
185
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #308539 |
|
Title:
|
Estimation of Covariogram Models by Minimum Prediction Contrast for Kriging
|
|
Author(s):
|
Marc Genton*+ and Olivier Perrin
|
|
Companies:
|
University of Geneva and University of Toulouse 1
|
|
Address:
|
Dept of Econometrics UNIGE, Geneva 4, CH-1211, Switzerland
|
|
Keywords:
|
Contrast ; Covariogram ; Kriging ; Prediction ; Space-time ; Wind speed
|
|
Abstract:
|
The method of spatial interpolation called kriging is based on the covariance structure of the data. Estimation of the parameters of a covariogram model are typically carried out by least squares fitting or by maximum likelihood, usually under normality for the latter case. We propose a new approach for the estimation of covariogram models based on the minimization of the integral of a contrast over the spatial domain of interest. The contrast is obtained from the variance of the difference between the kriging predictor based on the empirical covariogram and the one based on the parametric covariogram model. We derive properties of the resulting estimators from minimum contrast theory and perform various simulation studies to demonstrate the performances of our proposal. We apply our methodology to the estimation of space-time covariogram models for a dataset of Irish wind speed.
|
