ESTIMATION OF PARAMETERS OF PARAMETRIC/NON-PARAMETRIC SIMPLE LINEAR REGRESSION VIA SIMULATED DATA

Abstract

This study is on the estimation of parameters of Parametric/non-parametric simple linear regression via simulated data. Data of different sample sizes for both cases of residuals be normally distributed and non-normally distributed were simulated via “R Development”. was used in this study. The different data sets were tested for normality using Anderson-Darling technique. The algorithms for the parametric Theil’s and that of its non-parametric OLS regression were stated. It was concluded that the parametric OLS regression was better than its non-parametric Theil’s regression for both data whose residuals are normal and non-normal since their AIC and BIC are lower than that of Theil’s regression. Therefore the researchers recommended that future researchers should look at a similar work by comparing with other non parametric regression models, to know if the parametric OLS regression will still outperforms its non-parametric equivalents.