摘要：The aim of this paper was to compare, analyze and explain the change in linear, parabolic and cubic models of the relation between two chosen anthropometric soft tissue variables through a learning process implemented as a kinesiological treatment. A homogenous sample which homogenised further as the learning process advanced was used for the purpose of this research. In accordance with this, the linear model, the parabolic model and the cubic model which has never been used before depicting the relation between the thigh skinfold criterion variable and the upper arm circumference predictor variable were observed at 4 points in time. The convergence of statistically insignificant nonlinear models to statistically significant nonlinear model was observed and explained, providing a new methodological tool for quantitative analysis of the success of the learning process.The research has shown that the cubic model is the most appropriate one at the highest point of sample homogenization. It also suggests the introduction of a mathematical expression for an interval of numbers which encompasses those predictor variable values for which it is reasonable to calculate the criterion variable value – the domain of the (non) linear regression model. Furthermore, the process of calculating the extreme points of the general nonlinear model was mathematically described and applied.
