Table 3. Effect of neglecting phylogenetic uncertainty on estimates of divergence times under local clock models for the marsupial data.

split	MSS: MAP tree	MSS: 95% CS	HOPACH: MAP tree	HOPACH: 95% CS	UCLN (BEAST)
Sminthopsis/Phascogale	24.76	24.77	23.55	23.71	23.96
Sminthopsis/Echymipera	57.92	57.87	55.55	55.49	56.06
Echymipera/Perameles	12.39	12.39	11.09	11.16	13.09
Notoryctes/Sminthopsis	61.05	60.96	59.07	59.10	59.72
Dendrolagus/Pseudocheiridae	35.07	35.26	45.12	44.19	33.26
Dendrolagus/Phalanger	37.92	37.92	48.52	47.73	39.58
Phalanger/Vombatus	40.46	40.61	50.67	50.09	46.25
Vombatus/Phascolarctos	26.27	26.35	33.77	33.01	31.29
Vombatus/Dromiciops	55.80	56.52	62.54	62.43	61.46
Dasyurus/Rhyncholestes	73.23	72.20	76.74	74.30	77.67
Caenolestes/Rhyncholestes	10.66	10.49	13.88	11.75	13.71
Equus/Ceratomorpha	46.29	46.29	43.65	43.44	43.20
Cynocephalus/Leporidae	67.24	67.27	63.71	64.40	64.74
Cynocephalus/Ceratomorpha	80.23	80.25	76.66	76.81	77.54
Ceratomorpha/Bradypus	90.88	90.88	92.15	92.05	87.94
Bradypus/Sirenia	87.07	87.05	88.94	88.81	80.87
SSD	202.28	203.75	338.07	304.87	0.00

Results for local clock models, MSS and HOPACH, are given both for the maximum a posteriori (MAP) tree and as a posterior probability weighted average over the 95% credibility set (CS) of topologies sampled from the appropriate target distribution. The uncorrelated models from BEAST (UCLN: uncorrelated lognormal) integrate over tree topologies, branch lengths and parameters of the substitution model. Times are in million years ago. SSD are computed with respect to UCLN (baseline).