Multiple independent origins of auto-pollination in tropical orchids (Bulbophyllum) in light of the hypothesis of selfing as an evolutionary dead end

Table 4 Median maximum likelihood (ML) estimates of parameters, log-likelihoods and AIC values of alternative BiSSE models estimated across 1,000 Beast trees of Bulbophyllum clade C. Speciation rates are λ, extinction rates are μ, and character transition rates are q. Outcrossing is coded as ‘O’ and selfing as ‘S’

Model	d.f.	λ _O	λ _S	μ _O	μ _S	q _OS	q _SO	Ln L	AIC	ΔAIC	% trees¹ rejecting H₀
q _OS = 0	5	0.59	0.85	1.05	6.41E-09	0	0.76	-66.97	143.94	-21.83	100
λ _O = 0	5	0	1.54	1.32E-05	5.88E-08	2.92	5.82	-63.04	136.09	-13.98	98.5
λ _O = λ _S	5	0.66	0.66	1.26E-07	0.96	0.35	3.41E-06	-61.89	133.77	-11.67	99.3
Full ML	6	0.72	4.81E-07	5.21E-07	0.14	0.29	0.04	-58.68	129.35	-7.25	-
q _OS = q _SO	5	0.72	1.25E-06	3.68E-06	1.03E-05	0.30	0.30	-58.86	127.73	-5.62	0
q _SO = 0	5	0.71	4.54E-07	8.25E-07	0.25	0.28	0	-58.80	127.60	-5.49	0
μ _O = μ _S	5	0.71	2.15E-07	1.97E-06	1.97E-06	0.27	0.16	-58.75	127.50	-5.39	0
μ _S = 0	5	0.71	4.13E-07	6.50E-07	0	0.27	0.17	-58.75	127.50	-5.39	0
μ _O = 0	5	0.72	4.51E-08	0	0.14	0.29	0.04	-58.68	127.36	-5.25	0
λ _S = 0	5	0.72	0	1.32E-07	0.13	0.29	0.04	-58.68	127.35	-5.25	0
λ _S = 0, μ _O = 0	4	0.72	0	0	0.14	0.29	0.04	-58.68	125.35	-3.25	0
λ _S = 0, μ _O = 0, q _SO = 0	3	0.71	0	0	0.25	0.28	0	-58.80	123.60	-1.49	0
λ _S = 0, μ _O = 0, μ _S = 0, q _SO = 0	2	0.68	0	0	0	0.21	0	-59.05	122.11	0	0

For each model a “=” sign indicates constrained model parameters, while all other parameters were allowed to freely vary. D.f. = degrees of freedom, Ln L = log-likelihood, AIC = Akaike Information Criterion, ΔAIC = change in AIC relative to the best model. Models are sorted in ascending order of ΔAIC values. The best fitting model based on the AIC is in bold
*The final column shows the percentage of trees where the more complex (= full) model provided a significant improvement (P < 0.05) over the simpler constrained model (H₀) according to a likelihood-ratio test, where P-values were based on a χ ² distribution with degrees of freedom (d.f.) equal to the difference in the number of free parameters allowed by the two models (i.e., alternative [= full model] versus null [= constrained model] hypotheses)

ISSN: 2730-7182