Robust non-parametric regression models for some petroleum products

Abstract

The importance of statistics appears in trying of demonstrating different phenomena by models that are nearer to the reality. These models may be causative and are built on the basis of reason and result this is called Regression Model.  It has function shape and based on specific assumptions. But, sometimes we come up with a more flexible approach in case of absence of the knowledge of the studied phenomena or the first time made experiment or when we can't mention the causative function between the variables. This type of models is called Non parametric Regression. It is a type of regression in which the value of the independent variable doesn't take a specific shape, but built from information taken from the data and this requires a sample of volume bigger than the usual volume. For the parametric regression, the data suggest the structure of model and parameter estimation. Due to the importance of the petroleum products in lives of people that is considered as a source of civil and civilized development. Three petroleum products are taken (white oil, diesel oil, and fuel oil) through the non-parametric regression application. Besides, five non-parametric methods are taken (Lowes, Robust Lowes, Mean, Median, and Polynomial). They have been compared between each other, and we concluded that the robust non-parametric regression is the best way to compare the value of mean square error.