Ecological Systems and Devices Annotation << Back Some Peculiarities of Data Analysis With Interdependent Variables Panov V.G., Konstantinova E.D., Maslakova T.A., Kabakova E.A. Classical and modern methods of data analysis with correlated variables are discussed using heart rate variability (HRV) data analysis as an example. Some methods of selecting the most important variables and possible application of these variables for building a prognostic model are shown. The posed tasks are considered on the example of a database (DB) with anonymized data of 135 metallurgical plant employees. The possibility of describing a response (an HRV index) with the help of available variables is evaluated by constructing a linear regression model and calculating both usual and adjusted coefficients of determination. The values of the distance correlation coefficient dCor are calculated, which demonstrate the presence of non-linear relationships between some independent variables and the outcome and are not accounted for by the standard Pearson correlation coefficient. By constructing a random forest and calculating the Shepley values, the importance ranked lists of variables were obtained. The ranking by the above methods is robust to the presence of correlations, and allows us to highlight the most important variables by calculating their average rank considering both lists. It is also shown that the use of these variables for building prognostic models gives a good enough result. Keywords. Importance of variables, linear statistical models, correlation, distance correlation, coefficient of determination, corrected coefficient of determination, Shapley value, linear prognostic models. DOI: 10.25791/esip.5.2025.1521 Pp. 24-36.