Full Article: PDF
Scientific Object Identifier: http://s-o-i.org/1.1/TAS-06-122-46
DOI: https://dx.doi.org/10.15863/TAS.2023.06.122.46
Language: English
Citation: Udo, O. C., & Nwaogu, O. P. (2023). The Effect of Heterogeneity of Variance Data on Parametric and Nonparametric Regression Models. ISJ Theoretical & Applied Science, 06 (122), 282-288. Soi: http://s-o-i.org/1.1/TAS-06-122-46 Doi: https://dx.doi.org/10.15863/TAS.2023.06.122.46 |
Pages: 282-288
Published: 30.06.2023
Abstract: This study examined the effect of heterogeneity of variance data on parametric regression (OLS) and nonparametric regression (quantile regression- QR) models. The study was first subjected to heterogeneity of variance test via Breusch-Pagan-Godfrey technique, and it was revealed that there was existence of heterogeneity of variance in the data employed for the study. The multiple regression model of five explanatory variables, viz: shoulder width, elbow height, sitting height, arm length and age and the response variable (cholesterol) was first fitted with the adjusted coefficient of determination of 70.1% with the AIC being 26.245, as well as the quantile regression whose adjusted pseudo together with AIC are: , and for 25%, 50% and 75% respectively. The AIC agreed with the fact that the QR model was the best over the OLS model when there is presence of heterogeneity of variance in the data. The stepwise regression revealed that only three predictor variables (elbow height, age and shoulder width) were significantly related to the response variable at 5% level of significance. Comparison of parametric and nonparametric regression as the number of predictor variable increased to two and three also detected the presence of heterogeneity of variance, which gave QR advantage over OLS via their AIC values.
Key words: Parametric Regression, Nonparametric Regression, heterogeneity of variance, Pseudo-Values, AIC.
|
