Table 2.

Test data sets used to evaluate confidence interval performance. The last three columns record the average test error for a linear classifier trained using a training set of size n = 100 and loss function: squared error loss (SE), binomial deviance (BD), and ridged hinge loss (SVM)

Name	Features	Source	${\mathbb{E}}_{\tau }\left(\widehat{c}\right)\left(SE\right)$	${\mathbb{E}}_{\tau }\left(\widehat{c}\right)\left(BD\right)$	${\mathbb{E}}_{\tau }\left(\widehat{c}\right)\left(SVM\right)$
ThreePt	2	Simulated	.500	.500	.500
Quad	3	Simulated	.0997	.109	.101
Donut	3	Simulated	.235	.249	.232
Magic	11	UCI	.264	.231	.252
Mam.	6	UCI	.192	.190	.203
Ion.	9	UCI	.151	.147	.149
Bal.	5	UCI	.054	.050	.061
Liver	7	UCI	.342	.342	.334
Spam	10	UCI	.190	.183	.181
Heart	9	UCI	.167	.173	.174