A NONPARAMETRIC ROBUST ESTIMATOR FOR SLOPE OF LINEAR STRUCTURAL RELATIONSHIP MODEL

In this article, a nonparametric estimation procedure is proposed to estimate the slope of linear structural relationship model. Usually nonparametric methods are robust in nature, hence we propose a nonparametric method which is also robust and then it is compared with the traditional maximum likelihood method based on the normality assumption. For such a model, the maximum likelihood method is the best estimation method if no outlier exists in the data set. However, if the data contains outliers then sample estimates and results of maximum likelihood method could be unreliable. The real life example shows that our proposed method performs very well in estimating parameters and remains unaffected in the presence of outliers. The simulation study shows that in terms of mean square error our proposed estimator produces very satisfactory results in the presence of outliers.