Title: Bayesian Automating Fitting Functions for Spatial Predictions
Date: 1 July 2005
Authors: Monica Palaseanu-Lovejoy
Link: http://publications.epress.monash.edu/doi/pdf/10.2104/ag050014
Abstract:
A Bayesian predictive model for automating mapping of background radiation has the advantage of fully accounting for all uncertainties in the inferred data. Ten training datasets of background radiation were used to set up the model. The model is robust for data containing only close outliers but fails to accurately predict values when the input data is contaminated with extreme outliers, which are the result of a different random underlying process than the background data. For an integrated decision support system for automating mapping when data contamination is expected, a two stage approach is required in which background data are modeled with one set of equations and the contaminated data with a different set of equations.
Reference
Applied GIS
Volume 1, No. 2, August 2005
DOI: 10.2104/ag050014