Table 5.

Sensitivity of false positive rates to the choice of model parameterization under the nine different null scenarios of Simulation Studies 1–3

		SNR + DT cases				SNR (no DT)		High MNR (3a = DT)
	model	1a	1b	1c	3a	2a	2b	2a	2b	3a
M1a - M2a	GY	0.31	0.22	0.48	0.55	0.0	0.01	0.0	0.0	0.0
M1a - M2a	MG	0.25	0.41	0.91	0.94	0.0	0.15	0.0	0.0	0.0
M8ω = 1 - M8ω > 1	GY	0.47	0.24	0.58	0.56	0.0	0.02	0.0	0.0	0.0
M8ω = 1 - M8ω > 1	MG	0.44	0.57	0.94	0.97	0.0	0.18	0.0	0.0	0.0

Scenarios 1a and 1b are based on a 5-taxon tree, and 2a, 2b and 3a are based on a 17-taxon tree (see Fig. 1). GY denotes the frequency parameterization of Goldman and Yang [26] where the transition probability is proportional to target codon. MG denotes the frequency parameterization Muse and Gaut [47] where the transition probability is proportional to target nucleotide. Both require frequency estimates for the four nucleotides at each position of the codon (denoted F3 × 4), and thus each requires 9 free parameters