Table 1.

A simulation study to illustrate the percent computation time occupied by the relabeling algorithm for 1000 iterations of MCMC using a C++ extension of efficient CPU/GPU code for Bayesian analysis of truncated Dirichlet process mixture models (Suchard et al., 2010). This example uses data in p = 15 dimensions and examines running times for numbers of components k and sample sizes n that range across practically relevant values. All times are in seconds (real time).

n	k	Total Time	Relabeling	%
104	32	28.03	0.29	1.04
	64	53.83	1.46	2.71
	128	114.9	10.28	8.94
	256	287.4	78.87	27.4
105	32	77.82	0.94	1.21
	64	118.01	2.10	1.78
	128	206.80	10.93	5.28
	256	452.09	79.59	17.6
106	32	511.97	7.23	1.41
	64	734.13	8.29	1.13
	128	1156.16	17.16	1.48
	256	2100.11	85.78	4.08