Table 1.

Marginal log-likelihood values and posterior model probabilities in analyses of the two empirical datasets

Dataset	Introgression prob.	Log marginal likelihood	SE	P (model)
Adansonia	None	−1,701,076	3.35	4.9 × 10−28
	${\displaystyle {\phi }_{w\to x}}$	−1,701,050	3.47	9.5 × 10−25
	${\displaystyle {\phi }_{y\to z}}$	−1,700,996	3.49	0.27
	${\displaystyle {\phi }_{w\to x},{\phi }_{y\to z}}$	−1,700,995	3.55	0.73
Jaltomata	None	−1,914,473	3.33	6.8 × 10−46
	${\displaystyle {\phi }_{w\to x},{\phi }_{y\to z}}$	−1,914,404	3.18	6.3 × 10−16
	${\displaystyle {\phi }_{w\to x},{\phi }_{y\to z},{\phi }_{u\to v}}$	−1,914,369	3.53	1

Note: The log marginal likelihood was calculated using the stepping-stones sampling, with the standard error (SE) indicating uncertainty in the calculation. The assumed introgression probabilities in the model are defined in Supplementary Figs. S25 and S32 for the two datasets, respectively. The best model for each dataset is shown in bold.