A New Efficient Redescending M-Estimator for Robust Fitting of Linear Regression Models in the Presence of Outliers

Table 4

Total mean square error for contaminated datasets generated from normal distribution (90% X & Y∼N(0, 1) and 10% X∼N(20, 1); and 10% Y∼N(100, 1)).

N	P	OLS	Huber	Hampel	Andrew	Tukey	Ali
N = 30	P = 2	4.980	4.982	0.296	0.302	0.299	0.284
	P = 3	3.628	3.695	0.362	0.380	0.377	0.363
	P = 6	2.500	3.001	0.703	0.802	0.704	0.702
N = 50	P = 2	5.103	5.103	0.222	0.222	0.207	0.201
	P = 3	4.002	4.003	0.300	0.302	0.302	0.300
	P = 6	2.370	2.123	0.411	0.421	0.408	0.401
N = 100	P = 2	4.980	4.710	0.150	0.150	0.153	0.142
	P = 3	3.531	4.011	0.212	0.223	0.215	0.201
	P = 6	2.290	2.001	0.273	0.280	0.280	0.276