Table 2 . Runtime settings for IMa2 and STRUCTURAMA for the two real data sets: mouse (Geraldes et al. 2008) and chimpanzee (Fischer et al. 2006).

Data	La	Kb	nc	Programd	me	Burn-inf	Total	Samples	hng	Bmaxh	θmaxi	Tmaxj	Mmaxk
Mouse	7	2	113	IM tree	IS	1 × 105	1 × 106	5 × 105	40	0.85	10	5	0.5
Mouse	7	2	113	ST	NA	1 × 104	1 × 105	1 × 104	10	NA	NA	NA	NA
Chimp	10	4	39	IM island	IS	5 × 104	2 × 106	4 × 104	150	0.50	5	NA	1
Chimp	10	4	39	ST	NA	1 × 104	1 × 105	1 × 104	100	NA	NA	NA	NA
Chimp	9	2	20	ST	NA	1 × 104	1 × 105	1 × 104	10	NA	NA	NA	NA
Chimp	9	3	30	IM tree	IS	1 × 105	2 × 106	2 × 104	150	0.10	2	1.5	0.0001

^aNumber of loci.

^bNumber of populations: IMa2 was run with either a tree or an island structure of K populations, and STRUCTURAMA was run with a fixed number of K populations.

^cNumber of individuals.

^dIM tree, IMa2 with a population tree; IM island, IMa2 with an island; ST, STRUCTURAMA.

^eMutation model for IMa2. IS, infinite sites model; NA, not applicable.

^fBurn-in, total, and samples are in generations. After steps of burn-in, we take samples from total generations; one generation for IMa2 tries to update the population label of a single individual whereas one generation for STRUCTURAMA tries to update the labels of multiple individuals.

^gThe number of Metropolis-coupled chains (Geyer 1991).

^hThe heating level of the most heated chain.

ⁱMaximum of uniform prior of population size.

^jMaximum of uniform prior of splitting time.

^kMaximum of uniform prior of migration rates.