Comparison of the estimation methods used in estimating the nonparametric regression model with random errors subject to the ARMA model.

Abstract

This research dealt with comparing the methods used in estimating the nonparametric regression model with random errors subject to the (ARMA) model, as it was estimated by using the profile least squares method with (SCAD) and employing kernel smoothers for the local polynomial (Local Polynomial Smoother). ) When the degree of the polynomial is (P=1), we get the Local Linear Smoother, and (P=2), we get the Local Quadratic Smoother. The Bayesian Information Criterion, which is symbolized by the symbol (BIC), was also used to determine the rank of the error model. Likewise, the Mean Squared Error criterion, symbolized by the symbol (MSE), was used to compare the methods used in estimating the model. Realistic data was used by taking the data of the Asia Cell Communications Company through the trading of the company’s shares in the Iraqi Stock Exchange for the period from (1/3/2023 to 5/15/2023), and the most important thing that was achieved was the superiority of the positional squared regression method using Side least squares with SCAD having the lowest value for (MSE).