A Modified Maximum Likelihood Estimator for the Parameters of Linear Structural Relationship Model

The maximum likelihood method is the best method for estimating the parameters of a linear structural relationship model. However, if the data contains outliers then sample estimates and subsequent results of the maximum likelihood method could be unreliable. Robust estimation techniques have become popular due to its resistance to outliers. Thus, we proposed a modified maximum likelihood method to estimate the parameters of a linear structural relationship model where the non-robust components of the maximum likelihood methods are replaced by their corresponding robust alternatives. The simulation study and real life examples show that the proposed method performs very well in estimating the parameters in terms of estimated bias and mean square error.