Comparative Analysis of Three Linear Regression Methods Using Simulations

Juan Carlos Torres

 

 

Abstract

When it's necessary to fit a lineal model to a data set, least squares regression method is usually used. This method is optimum if the residuals distribution is normal. When the assumption of residuals normality doesn't comply it's necessary to use alternative regression methods, as Least absolute deviations (LAD) or Non parametric regression based on ranks, which don't need the assumption about the residuals distribution and allow a better estimation of regression model parameters.

 

Key words: Linear Model, Regression Analysis, Simulation.

 

PDF  (Spanish)