Ecological Systems and Devices Annotation << Back Regression Models in Medico-Ecological Monitoring: an Algorithm for Taking Into Account Correlations Between Predictors Varaksin A.N. Regression models in medico-ecological monitoring (for example, in problems of the relationship between population morbidity Y and environmental pollution indicators X) may contain interrelated predictors characterizing health risks (correlations between predictors X). The presence of correlations between predictors in regression models leads to diffi culties in interpreting the results of regression modeling. Currently, there is no clear solution to the problem of correlated predictors in regression models. The purpose of the study is to develop one of the possible algorithms for solving this problem. The algorithm is implemented in two stages. At the fi rst stage, a regression model containing so-called “cross terms” is constructed using conventional methods of regression analysis. Such a model allows, in principle, to calculate the effects of some predictors X on the response Y for different values of other predictors, which is a prerequisite for taking into account correlations between predictors. Such a model, however, does not contain specifi c indicators of the correlation of predictors, for example, correlation coeffi cients between them, and does not allow correlations to be “automatically” taken into account when analyzing modeling results. Therefore, at the second stage of implementing the algorithm, it is proposed to use either explicit statistical relationships between predictors or moving average methods to explicitly take into account correlations between predictors. The connection between the general morbidity of the population of St. Petersburg and atmospheric air pollution with carbon monoxide and sulfur dioxide is considered in a specifi c example. This example shows the details of the implementation of the algorithm for taking into account correlations between predictors. The results of the implementation of the algorithm are compared with the results of existing approaches to the analysis of regression models, the advantages of the new algorithm are shown. Keywords: regression models, correlated predictors, cross terms, moving average, health risks, overall disease prevalence, air pollution, carbon oxide, sulfur dioxide. DOI: 10.25791/esip.1.2025.1495 Pp. 03-14.