Constructing the nonlinear regression equations based on multivariate normalizing transformations

Abstract

In the paper we consider the techniques to construct the equations, confidence and prediction intervals of nonlinear regressions on the basis of multivariate normalizing transformations for non-Gaussian data. We demonstrate that the poor normalization of multivariate non-Gaussian data using the univariate transformations leads to an expansion of the confidence and prediction intervals of non-linear regression for a larger number of data rows compared to the multivariate normalizing transformation.