Table 2.

Analysis of Reference Data Sets.^a

			BUSTED				BUSTED[S]						Sites
Gene	S	N	P Value	$\text{CV}(\omega )$	ω 3	p 3	P Value	$\text{CV}(\omega )$	$\text{CV}(\alpha )$	ω 3	p 3	α Distribution	ΔAICc	$++$	$+-$	$-+$
β-Globin	17	144	$<{10}^{-4}$	4.66	20.54	2.7%	$<{10}^{-4}$	3.96	1.46	9.60	3.4%	0.42 (53%), 1.3 (44%), 6.5 (3%)	37.13	4	2	0
Flavivirus NS5	18	342	0.42	4.17	1.14	4.4%	0.49	4.60	1.71	1.11	2%	0.12 (19%), 0.53 (72%), 6.2 (9%)	259.28	0	0	0
Primate COXI	21	510	0.5	5.60	1.00	2.5%	0.5	6.32	2.35	1.00	1.1%	0.04 (2%), 0.58 (95%), 13.5 (3%)	105.59	0	0	0
Drosophila adh	23	254	0.0003	4.73	4.78	2.6%	0.0016	4.62	0.59	4.26	2.4%	0.53 (40%), 1.0 (52%), 3.0 (8%)	18.02	1	0	0
Encephalitis env	23	500	0.5	0.0	1.00	0%	0.5	0.0	0.67	1.14	0%	0.41 (31%), 1.1 (66%), 4.6 (3%)	46.32	0	0	0
Sperm lysin	25	134	$<{10}^{-4}$	2.45	17.13	10%	$<{10}^{-4}$	2.39	0.87	17.38	7.6%	0.21 (38%), 1.1 (46%), 2.5 (16%)	160.49	23	2	0
HIV-1 vif	29	192	0.0002	1.68	3.14	26%	0.025	18.28	1.06	988.84	0.05%	0.30 (54%), 1.2 (34%), 3.6 (12%)	188.31	0	1	0
Hepatitis D virus antigen	33	196	$<{10}^{-4}$	3.81	16.42	2.9%	$<{10}^{-4}$	3.58	0.92	16.61	1.9%	0.05 (22%), 0.78 (58%), 2.7 (20%)	273.38	8	2	0
Vertebrate Rhodopsin	38	330	$<{10}^{-4}$	6.78	20.76	1%	$<{10}^{-4}$	5.57	1.42	7.06	1.0%	0.36 (57%), 1.1 (38%), 8.2 (5%)	473.77	5	3	0
Influenza A virus HA	86	329	0.5	2.14	1.00	27%	0.095	1.11	0.85	2.10	18%	0.49 (62%), 1.3 (29%), 3.4 (9%)	130.9	0	0	0
Camelid VHH	212	96	$<{10}^{-4}$	3.22	28.51	3.9%	$<{10}^{-4}$	3.05	0.84	26.87	1.9%	0.24 (33%), 0.85 (45%), 2.5 (22%)	1,436.46	26	11	1

^aResults from the reanalysis of the data sets used in Kosakovsky Pond and Muse (2005), Yokoyama et al. (2008), and Chen and Sun (2011), arranged by sequence count. We ran selection tests with three nonsynonymous and three synonymous rate categories (for BUSTED[S]). Column headings are as follows: S, number of sequences; N, number of codons; $\text{CV}(\omega )$, the coefficient of variation (CV) for the inferred distribution of ω ratios; ω₃, the maximum likelihood estimate (MLE) of the strength of selection; p₃, the MLE of the proportion of sites under selection (proportion of sites in the ω₃ category); $\text{CV}(\alpha )$, the CV for the inferred distribution of synonymous rates; ΔAIC_c, the difference between AIC_c values of BUSTED and BUSTED[S]. The α distribution columns list the estimated values of the 3 α categories along with their estimated frequencies. The Sites columns count the number of alignments where at least one method called a site selected (using evidence ratio of at least 5): $++$ both methods yes; $+-$ BUSTED yes, BUSTED[S] no; $-+$ BUSTED no, BUSTED[S] yes.