Background. When processing the results of observational studies we ave to usemultivariate statistical methods that will examine the simultaneous influence of both independent and confounding variables on the outcome, Objective: The aim of this paper is to further explain how a researcher could decide whether multiple linear regression is suitable statistical option for processing his (her) data, and then how to implement it properly. Methods: This article is a narrative review of literature about logic, assumptions, quality check and interpretation of multiple linear regression. Results: Multiple linear regression is a complex linear equation (model) in which on one side of the equal sign is the absolute value of the dependent variable (i.e., the outcome), and on the other is a sum of additions, of which only one is a constant, and all others are the product of an independent or confounding variable and their coefficients. After checking assumptions and quality of the model, we may decide whether a predictor has significant influence on outcome, or not, and calculate size of this influence. Conclusions: Multiple linear regression is an extremely useful statistical model for explaining the influence of multiple predictors simultaneously on a continuous type dependent variable, but it requires the fulfillment of fairly strict assumptions in order to be used. That's why multiple linear regression should be used only when the conditions are met, otherwise other types of linear and non-linear models whose assumptions are far more lenient should be resorted to. .
Key words: multiple linear regression; assumptions; heteroscedascity; collinearity
|
