Table 7: Prediction mean squared error with ten observations on $n_{s} = 200$ test samples.

Test samples $y_{i}^{(8)}$ and $y_{i}^{(9)}$ and $y_{i}^{(10)}$ are treated as the response and the rest of the observations are used as the training data to estimate parameters used to predict the response. Prediction accuracy is measured by mean squared error (MSE) between simulated responses and predicted responses. Values presented are the mean MSE (Err) and standard deviation (SD) across 20 runs of each method. Standard deviation (SD) is missing for SCCA because the method is deterministic.

		BASS
	$n_{t}$	EM		MCMC-EM		PX-EM		sGFA		GFA		SCCA	MRD-lin
	$n_{t}$	Err	SD	Err	SD	Err	SD	Err	SD	Err	SD	Err	Err	SD
Sim5	10	1.01	0.020	1.00	0.011	1.00	0.007	0.99	0.008	1.00	0.002	0.99	1.49	0.001
	30	0.88	0.031	0.86	0.018	0.87	0.028	0.89	0.005	0.90	0.002	0.99	1.01	0.035
	50	0.86	0.023	0.85	<1e-3	0.86	0.022	0.87	0.003	0.88	0.001	0.99	0.97	0.020
	100	0.85	0.007	0.85	<1e-3	0.85	0.002	0.86	0.003	0.87	0.001	1.01	0.92	0.039
	200	0.85	0.006	0.84	<1e-3	0.84	<1e-3	0.84	0.001	0.83	0.001	0.96	1.06	0.105

Sim6	10	0.61	0.164	0.57	0.116	0.51	0.031	0.58	0.012	0.75	0.011	0.97	1.00	<1e-3
	30	0.49	0.160	0.40	0.093	0.38	0.007	0.43	0.006	0.40	0.005	0.98	0.46	0.006
	50	0.44	0.099	0.39	0.011	0.39	0.004	0.41	0.002	0.40	0.001	1.01	0.42	0.009
	100	0.39	0.033	0.39	0.004	0.39	0.011	0.39	0.002	0.39	0.001	0.97	0.52	0.249
	200	0.38	0.003	0.38	0.001	0.38	0.001	0.39	0.001	0.39	0.001	1.01	0.40	0.020

Table 7: Prediction mean squared error with ten observations on ns=200 test samples.

Table 7: Prediction mean squared error with ten observations on $n_{s} = 200$ test samples.