. 2017 May 3;15:264–273. doi: 10.1016/j.nicl.2017.04.027

Table 5.

Regression modelling performance measures using individual MMN responses to the left gap stimulus paradigm with 0–450 ms temporal mask, mapped to participants' GAF scores. p-values are Sidák corrected for multiple comparisons.

Algorithm	Normalisation	Correlation	R p-value uncorrected	R p-value Šidák corr.	R²	R²p-value uncorrected	R²p-value Šidák corr.	MSE	Norm. MSE	MSE p-value uncorrected	MSE p-value Šidák corr.
Kernel Ridge Regression	None	0.59	0.0005	0.0030	0.35	0.0010	0.0060	342.42	5.90	0.0107	0.0625
Kernel Ridge Regression	Normalised	0.74	0.0001	0.0006	0.55	0.0001	0.0006	195.20	3.37	0.0001	0.0006
Relevance Vector Regression	None	0.59	0.0002	0.0012	0.35	0.0006	0.0036	342.48	5.90	0.0371	0.2029
Relevance Vector Regression	Normalised	0.73	0.0001	0.0006	0.53	0.0001	0.0006	156.08	2.69	0.0001	0.0006
Gaussian Process Regression	None	0.59	0.0002	0.0012	0.35	0.0005	0.0030	342.43	5.90	0.0116	0.0676
Gaussian Process Regression	Normalised	0.73	0.0001	0.0006	0.53	0.0001	0.0006	156.33	2.70	0.0001	0.0006