On two flexible methods of 2-dimensional regression analysis
Loading...
Date
2012
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Technická univerzita v Liberci, Česká republika
Abstract
The paper deals with the problem of non-parametric statistical modeling of 2-dimensional sur- faces from observed data, i.e. the regression analysis. In general, the model is constructed from a set of basal functions, as are the splines, gaussians and others. However, such model- ing means to estimate a large set of parameters (locations of functional units and parameters of their combination). We shall present two approaches allowing reduction of the number of needed parameters. Namely, a well known method of projection pursuit, and the less known method of Gordon surface. Further, we shall analyze possible serious consequences of sparse data to precision of model and uncertainty of prediction. Methods will be illustrated in artificial examples.
Description
Subject(s)
statistics, regression analysis, splines, projection pursuit, Gordon surface, prediction error
Citation
Item identifier
ISSN
1803-9782