Ecological Systems and Devices Annotation << Back APPLICATION OF MOVING AVERAGE METHODS FOR THE CONSTRUCTION OF REGRESSION MODELS IN MEDICAL AND ENVIRONMENTAL RESEARCH A.N. Varaksin, Yu.V. Shalaumova, T.A. Maslakova, G.M. Nasybullina, S.S. Delets In this article, the moving average method was used to construct multiple linear regression (MLR) models. The method was most effective in the case of a large number (several tens) of potential predictors X, each of which has low pair correlation coefficients with a response Y. The moving average method allowed to analyze the structure of relations Y with each X and to visually assess the improvements in predicting the response values Y that appear when another predictor is added to the MLR. We have proposed a way to create a database of moving averages through which multiple regression models can be built. Examples of using the moving average method in the field of medical and environmental research are given. When using moving average methods, the same response Y can be described by MLR models with a different set of predictors of approximately equal quality (with the same determination coefficient R2). We have analyzed MLR models that describe one of the indicators of the physical development of adolescents (body mass index), from the point of view of the subject area (medical and environmental). MLR based on the moving average method allows us to identify such relationships between the body mass index and its predictors that cannot be detected using primary data due to the small pair correlation coefficients between the response Y and the predictors X. Keywords: moving average method; multiple linear regression; adolescents; body mass index; nutrition; physical activity. DOI: 10.25791/esip.06.2020.1159 Pp. 12-21.