Table 3. GBTM selection No. of groups .... | American Society of Hematology

Table 3.

GBTM selection

No. of groups	Group sizes	Probability of group assignment	BIC lower confidence limit∗	Bootstrap bias-corrected BIC	BIC upper confidence limit∗	Bayes factor/2log_e(B10)†
1			–4275.24	–4187.38	–4099.52
	343 (100%)	1.0
2			–4171.36	–4092.35	–4013.34	190.06
	236 (69.3%)	0.96
	107 (30.7%)	0.91
3			–4091.61	–4017.84	–3944.07	149.02
	108 (32.6%)	0.89
	199 (56.5%)	0.90
	36 (10.9%)	0.93
4			–4087.75	–4013.95	–3940.15	7.79
	8 (2.4%)	0.98
	141 (40.8%)	0.88
	153 (45.0%)	0.88
	41 (11.8%)	0.92
5‡			–4120.00	–4040.92	–3961.84	–53.95
	8 (2.4%)	0.98
	130 (38.4%)	0.89
	160 (45.3%)	0.87
	36 (11.3%)	0.93
	9 (2.6%)	0.99

No. of groups	Group sizes	Probability of group assignment	BIC lower confidence limit∗	Bootstrap bias-corrected BIC	BIC upper confidence limit∗	Bayes factor/2log_e(B10)†
1			–4275.24	–4187.38	–4099.52
	343 (100%)	1.0
2			–4171.36	–4092.35	–4013.34	190.06
	236 (69.3%)	0.96
	107 (30.7%)	0.91
3			–4091.61	–4017.84	–3944.07	149.02
	108 (32.6%)	0.89
	199 (56.5%)	0.90
	36 (10.9%)	0.93
4			–4087.75	–4013.95	–3940.15	7.79
	8 (2.4%)	0.98
	141 (40.8%)	0.88
	153 (45.0%)	0.88
	41 (11.8%)	0.92
5‡			–4120.00	–4040.92	–3961.84	–53.95
	8 (2.4%)	0.98
	130 (38.4%)	0.89
	160 (45.3%)	0.87
	36 (11.3%)	0.93
	9 (2.6%)	0.99

BIC, Bayesian Information Criterion.

Confidence limits for bootstrapped BIC reflect 1 SD from the mean.

Difference between model with N groups and model with N-1 groups.

Singular convergence error required respecifying group 3 with linear trajectory shape.