Table 2 . (A) Power to discriminate between alternative demographic models using an “out-of-bag” procedure given the parameter model specification.

A.
Classified models Simulated models	A2	B3	C2	D8	E1	E2	E3	E4	E5	Total	Classification error	Probability that E4 is selected
A2	8902	1043	24	12	2	9	0	1	7	10000	11.0%	0.0%
B3	2151	7616	26	15	4	73	11	25	79	10000	23.8%	0.3%
C2	210	463	4844	1576	675	642	534	469	587	10000	51.6%	5.5%
D8	447	330	3067	1905	1039	662	908	911	731	10000	81.0%	10.7%
E1	355	336	2094	1342	1462	735	1323	1339	1014	10000	85.4%	15.8%
E2	400	1317	2161	1039	643	1473	756	1062	1149	10000	85.3%	12.5%
E3	28	625	1894	1173	1277	857	1641	1438	1067	10000	83.6%	17.0%
E4	344	1096	1024	993	1219	918	1313	2009	1084	10000	79.9%	23.7%
E5	435	1291	1491	777	963	1170	1006	1228	1639	10000	83.6%	14.5%
Total	13272	14117	16625	8832	7284	6539	7492	8482	7357

The comparisons are made at the final selection step between the most likely one- to four-wave introduction models and all possible five-wave introduction models. The table shows how many of the 10,000 simulated datasets generated under a given scenario (A2 to E5, rows) were classified into each demographic scenario (A2 to E5 columns). The number of incorrect classifications is then used to compute the overall classification error. The last column shows the percentage of simulated models classified as E4 (which was the most likely scenario for the observed genetic dataset). Bold numbers indicate correct classification, and underlined numbers indicate >10% incorrect classification. (B) Probability of a given number of origins given that the E4 model is selected.

Quick links

Search