Problem multikolinearnosti u višestrukoj linearnoj regresiji / Problem of multicollinearity in a multiple linear regression
2016
Novaković, Tihomir (Poljoprivredni fakultet, Novi Sad (Srbija)) | Nikolić-Đorić, Emilija (Poljoprivredni fakultet, Novi Sad (Srbija)) | Mutavdžić, Beba (Poljoprivredni fakultet, Novi Sad (Srbija))
The aim of this paper is to consider the problem of multicollinearity in multiple linear regression. Multicollinearity in the regression model includes the presence of full (extreme multicollinearity) or approximate linear correlation of independent variables. There are numerous analytical methods which can be used for the observation of this problem. Also, the paper discusses causes and consequences of multicollinearity and gives specific recommendations on how to overcome it. For the illustration of multicollinearity, we used a real data series established by experimental trials in the period 1997-2001. Wheat yield was taken as a dependent variable, while 24 numerical indicators relevant for monitoring the development of the plant were used as predictor variables. Calculations were performed using the programs R 3.3.2, STATISTICA 13 and STATA 13. There was a discrepancy of results of the applied programs which indicates a problem in calculating the numerical evaluation of parameters, possibly resulting from multicollinearity.
Mostrar más [+] Menos [-]Palabras clave de AGROVOC
Información bibliográfica
Este registro bibliográfico ha sido proporcionado por Matica Srpska Library