Table 2.

The predictive performance of the longitudinal regression tree algorithms with and without the autocorrelation structure between within-subject errors in the simulated dataset

Longitudinal regression tree algorithm	Autocorrelation structure	MSE	MAD	Deviance
RE-EM		0.3727947	0.4524195	28636.44
	AR (1)	24.52032	3.943993	29912.64
	CS	0.3782275	0.4584203	28750.63
RE-EM Unbiased		0.2657871	0.4305183	25739.56
	AR (1)	17.03199	2.813750	27807.73
	CS	0.2669871	0.4354071	25946.75
Ev-RE-EM		0.2735940	0.4317192	25847.34
	AR (1)	17.57367	2.835460	27821.97
	CS	0.2747831	0.4384521	25997.92

: variance-covariance diagonal matrix of errors, AR (1): first-order autoregressive process, CS: compound symmetry structure with a constant correlation (unstructured)