Table 1. Performance of methods as a function of sampling strategy and mean migration rate in the two-population “fixed tree” scenario.

Sampling	Rate c	Method	Calibration d	Correlation e	RMSE f
Even a	Fast	DTA	0.56	0.58	1.83
		MTT	0.87	0.77	1.32
		BASTA	0.95	0.83	1.51
Even	Slow	DTA	0.81	0.64	1.65
		MTT	0.96	0.75	1.52
		BASTA	0.97	0.81	1.30
Uneven b	Fast	DTA	0.68	0.33	1.79
		MTT	0.80	0.46	2.50
		BASTA	0.84	0.70	2.08
Uneven	Slow	DTA	0.80	0.39	1.73
		MTT	0.85	0.42	2.49
		BASTA	0.88	0.51	2.29

For each combination of sampling strategy, migration rate and method, we assessed the methods’ performance across 100 replicates by recording the “true” (i.e. simulated) ratio of the migration rates f _1,2/f _2,1, the point estimate (posterior median) and the 95% credible interval.

^a 100 samples per population.

^b 10 samples for one population and 190 for the other.

^c total mean migration rate: fast ($\bar{f}=5.0$) or slow ($\bar{f}=0.5$).

^d proportion of replicates for which the truth fell within the 95% credible interval.

^e correlation between the truth and the point estimate.

^f root mean square error of the point estimate.