Skip to main content

Table 1 Coverage summary for simulation study

From: bModelTest: Bayesian phylogenetic site model averaging and model comparison

 

Site

Rate coverage

Mean

Subst. Model

Freqs

model

AC

AG

AT

CG

CT

GT

rate

coverage

Equal

plain

93

97

94

96

95

95

95

98

Equal

+G

91

95

93

93

95

93

93.3

97

Equal

+I

92

94

94

95

93

94

93.6

96

Equal

+G+I

89

96

95

94

95

95

94

98

Unequal

plain

96

95

96

97

93

96

95.5

96

Unequal

+G

95

94

94

94

96

96

94.8

98

Unequal

+I

89

94

95

95

93

95

93.5

93

Unequal

+G+I

97

94

94

93

93

96

94.5

97

Mean

 

94.25

94.25

94.75

94.75

93.75

95.75

94.6

96

 

Site

Site model

  

Frequency

Frequency coverage

Freqs

model

coverage

α

p inv

coverage

A

C

G

T

Equal

plain

100

  

100

100

100

100

100

Equal

+G

96

94

 

100

100

100

100

100

Equal

+I

98

 

95

100

100

100

100

100

Equal

+G+I

99

89

88

100

100

100

100

100

Unequal

plain

100

  

100

92

95

97

96

Unequal

+G

97

94

 

100

97

92

92

98

Unequal

+I

98

 

92

100

95

94

94

89

Unequal

+G+I

100

93

91

100

99

96

96

98

Mean

 

98.75

93.50

91.50

100.00

97.38

97.88

97.13

97.38

  1. The first column lists the frequency and site models used to generate the data, and the last row is the mean coverage over all 800 runs. Coverage for rate parameters and frequencies is defined as the number of replicate simulations in which the true parameter value was contained in the estimated 95% HPD interval. The mean rate column contains the coverage averaged over all six rate coverage columns (i.e. the proportion of the 600 parameter estimates whose values were contained in their respective 95% HPD intervals. For details of substitution model coverage see text. The site model coverage is the number of replicate simulations that contained the correct model specification for rate heterogeneity across sites in the 95% credible set of models. Columns α and p inv are coverages of the shape and proportion invariable parameter conditioned on sampling from the true site model