Nonlinear regression modeling on the base of Monte-Carlo method

An effective method of nonlinear regression modeling is proposed, which is based on direct minimization of the sum of the squares of the differences by means of Monte-Carlo procedures. Such approach allow the application for nonlinear modeling practically any elementary functions and theirs compositions. The selection of one-dimensional function for real experimental data is realized by means of the graph-analytical method. The multidimensional regression function is constructed by means of the consistent transition from one-dimensional model to two-dimensional model and further by transforming the model coefficients of the proceeded level in the functions of the additionally accounted factor. New possibilities for the nonlinear regression analyses are demonstrated on three real examples from sphere of the mycology, plant biology and chemical technology.