Table 1.

Regularization parameter estimates found for SVR, ridge regression and PLS. With the exception of Ridge(HKB), all parameters were estimated in the secondary CV and these are the average parameter values over the 500 iterations for each simulation setup.

	U(0,4)	U(2,6)	U(4,8)	U(6,10)
	SVR-ε
noise σ = 0.8	0.466	0.503	0.494	0.483
noise σ = 1.6	0.537	0.562	0.539	0.565
noise σ = 3	0.595	0.631	0.659	0.661
	SVR-C
noise σ = 0.8	2.840	0.496	0.318	0.277
noise σ = 1.6	0.708	0.195	0.159	0.149
noise σ= 3	0.301	0.134	0.120	0.118
	Ridge(λ-HKB)
noise σ = 0.8	10.188	17.734	23.315	28.990
noise σ = 1.6	22.914	49.834	66.571	83.993
noise σ = 3	33.320	79.759	114.133	148.313
	Ridge(λ-CV)
noise σ = 0.8	14.992	1.608	0.310	0.000
noise σ = 1.6	39.201	18.395	6.151	0.698
noise σ = 3	40.000	39.441	32.404	18.840
	PLS-component number
noise σ = 0.8	8.095	12.314	8.709	6.807
noise σ = 1.6	2.051	8.009	8.268	7.635
noise σ= 3	2.000	3.638	4.683	6.013