BMC Medical Research Methodology

Table 1 Outlier identification results on simulated data with p = 15

From: Outlier detection in spatial error models using modified thresholding-based iterative procedure for outlier detection approach

	Outlier = 50			Outlier = 20			Outliers = 10
	M	S	JD	M	S	JD	M	S	JD
No leverage
Spatial-hard-IPOD	0.0036	0.0533	87	0.0090	0.0484	88	0.0400	0.0459	76
Spatial-soft-IPOD	0.0030	0.0518	86	0.0145	0.0195	74	0.0520	0.0129	52
Hard-IPOD	0.1500	0.0038	65	0.2600	0.0054	54	0.2430	0.0132	50
Soft-IPOD	0.3296	0.0072	65	0.3615	0.0016	55	0.3750	0.0004	56
MM	0.4192	0.0110	40	0.4805	0.0188	37	0.4800	0.0335	37
LTS	0.1914	0.0076	49	0.2970	0.0148	48	0.2680	0.0236	45
GY	0.0714	0.0680	0	0.1490	0.1120	0	0.1210	0.1221	2
Leverage = 15
Spatial-hard-IPOD	0.0027	0.0505	90	0.0095	0.0439	87	0.0410	0.0445	72
Spatial-soft-IPOD	0.0016	0.0550	92	0.0090	0.0258	83	0.0500	0.0123	57
Hard-IPOD	0.2714	0.0041	63	0.2235	0.0058	59	0.2050	0.0058	60
Soft-IPOD	0.7059	0.0124	29	0.4365	0.0024	50	0.3650	0.0010	59
MM	0.5137	0.0117	37	0.5010	0.0259	36	0.3960	0.0358	50
LTS	0.2804	0.0108	59	0.2720	0.0162	56	0.2300	0.0220	64
GY	0.1310	0.0740	0	0.1355	0.1057	0	0.1250	0.1234	1
Leverage = 20
Spatial-hard-IPOD	0.0046	0.0487	85	0.0110	0.0437	87	0.0330	0.0457	79
Spatial-soft-IPOD	0.0031	0.0542	85	0.0115	0.0238	80	0.0540	0.0127	57
Hard-IPOD	0.2600	0.0044	69	0.2325	0.0074	62	0.3250	0.0021	56
Soft-IPOD	0.6923	0.0207	31	0.4240	0.0032	51	0.5020	0.0008	48
MM	0.3769	0.0166	46	0.4525	0.0215	40	0.4370	0.0384	47
LTS	0.1831	0.0055	69	0.2340	0.0169	62	0.3810	0.0174	47
GY	0.1015	0.0699	0	0.1060	0.1138	88	0.1930	0.1213	3

Seven methods are compared: Our proposed hard-IPOD, our proposed hard-IPOD, hard-IPOD, soft-IPOD, MM-estimator, LTS and Gervini–Yohai’s fully efficient one-step procedure

Back to article page

ISSN: 1471-2288

Contact us

General enquiries: journalsubmissions@springernature.com