TABLE 2.

Comparison with other methods

Parameters	Method	γ	ρ	f
γ = 1, ρ = 1 (f = 1)	SummStat	0.58	0.76	0.49
	MaxHap	0.63	0.89	0.52
	GenCo	0.77	0.96	0.63
γ = 10, ρ = 1 (f = 10)	SummStat	0.83	0.77	0.57
	MaxHap	0.99	0.81	0.77
	GenCo	1	0.94	0.9

For 1000 data sets simulated with ρ = 1 and γ ∈ (1, 10), we compare our results (GenCo) with those from maxHap (Hudson 2001), and for 100 of the same data sets we also show results obtained using a third method (summStat) based on that of Padhukasahasram et al. (2006) (see text). MaxHap was run over a grid of 11 f values ranging from 0 to 2.5 or 25 (inclusive) for the simulated data sets with f = 1 and f = 10, respectively. SummStat was run on a coarse grid of ρ ∈ (8, 10, 20, 40, 45), and γ ∈ (8, 10, 20, 40, 45) for the first test and γ ∈ (80, 100, 200, 400, 450) for the second test. For each parameter γ, ρ, and f we present the proportion of data sets for which the estimated value was within a factor of 2 of the value used to simulate data.