Estimating misclassification error: a closer look at cross-validation based methods

BMC Research Notes

Table 3 Simulation results for p = 5, Σ₁ = Σ₂ = Ι₍₅₎, N = 1000

Method	n	_Δ	ē ^a	ē _N	$\overset{―}{M S E}$	SD(MSE)	$\overset{―}{V A R}$	SD(VAR)	MSB	$\overset{―}{B I A S}$	SD(BIAS)
LOOCV	50	1	0.38308	0.38848	0.00688	0.00996	0	0	0.00688	0.00540	0.08281
	50	3	0.09741	0.10052	0.00227	0.00338	0	0	0.00227	0.00311	0.04760
BCVn	50	1	0.38308	0.27058	0.02073	0.01027	0.00587	0.00162	0.01486	−0.11250	0.04693
	50	3	0.09741	0.07686	0.00305	0.00167	0.00160	0.00082	0.00145	−0.02055	0.03208
LOOCV	100	1	0.35366	0.35164	0.00291	0.00419	0	0	0.00291	−0.00202	0.05390
	100	3	0.07962	0.07906	0.00080	0.00110	0	0	0.00080	−0.00056	0.02822
BCVn	100	1	0.35366	0.29312	0.00799	0.00454	0.00308	0.00074	0.00491	−0.06054	0.03529
	100	3	0.07962	0.06781	0.00141	0.00080	0.00075	0.00031	0.00066	−0.01180	0.02278
k CV10	50	1	0.38308	0.39236	0.00711	0.00860	0.00126	0.00036	0.00585	0.00929	0.07598
	50	3	0.09741	0.10438	0.00252	0.00327	0.00039	0.00019	0.00213	0.00697	0.04568
BCV10	50	1	0.38308	0.27549	0.01950	0.00968	0.00579	0.00113	0.01371	−0.10758	0.04622
	50	3	0.09741	0.08135	0.00302	0.00163	0.00173	0.00078	0.00128	−0.01606	0.03202
k CV10	100	1	0.35366	0.35585	0.00299	0.00364	0.00049	0.00012	0.00251	0.00219	0.05003
	100	3	0.07962	0.08055	0.00085	0.00109	0.00010	0.00005	0.00075	0.00094	0.02735
BCV10	100	1	0.35366	0.29629	0.00756	0.00427	0.00305	0.00041	0.00451	−0.05737	0.03488
	100	3	0.07962	0.06942	0.00136	0.00072	0.00076	0.00026	0.00060	−0.01020	0.02239

$\overset{―}{e} a = (1 / N) \sum_{i = 1}^{N} e_{i}, {\overset{―}{e}}_{N} = (1 / N) \sum_{i = 1}^{N} {\overset{―}{e}}_{Ri}, M S B = (1 / N) \sum_{i = 1}^{N} {({\overset{―}{e}}_{Ri} - e_{i})}^{2}$

ISSN: 1756-0500