Table 2.

Generalizability of behavioural predictions based on clustering-defined groups.

	CALM	%ile	train	test	p	ACE	%ile	train	test	p
Matrix Reasoning		50	0.62	0.74	0.003		50	0.65	0.71	0.027
Vocabulary		65	0.68	0.64	0.028		55	0.89	0.72	0.025
Digit Recall		65	0.60	0.56	0.214		60	0.61	0.67	0.089
Dot Matrix		50	0.63	0.56	0.214		70	0.61	0.50	0.387
Backward Digit Recall		50	0.62	0.58	0.214		50	0.71	0.77	0.025
Mr. X		50	0.60	0.69	0.004		70	0.5	0.71	0.025
Delayed Story Recall		50	0.62	0.64	0.028		N/A	N/A	N/A	N/A

The optimal percentile cut-off identified in the training data set and the training performance are listed in the columns %ile and train. The fraction of participants who were accurately predicted to score above or below the median of all cases based on their group assignment is listed for the train and test set column. A statistical comparison to random assignment (1000 permutations) is listed in the column labelled p.