Table 1.
Type of selection | Model | Pre-epidemic | Post-epidemic | Post-sex | n | AIC | evidence ratio | |||
---|---|---|---|---|---|---|---|---|---|---|
No selection | 1) unimodal: null |
μA σA |
μA σA |
μA σA |
2 | -45.29 | 2145 | |||
2) bimodal: null |
μ1A σA |
μ2A σA |
μ1A σA |
μ2A σA |
μ1A σA |
μ2A σA |
3 | -51.61 | 90.8 | |
Directional selection | 3*) unimodal: mean, var differ |
μA σA |
μB σB |
μC σC |
6 | -44.74 | 2817 | |||
4) unimodal: pre-epidemic differs |
μA σA |
μB σB |
μB σB |
4 | -47.99 | 557 | ||||
5*) bimodal: all modes differ |
μ1A σA |
μ2A σA |
μ1B σA |
μ2B σA |
μ1C σA |
μ2C σA |
7 | -56.96 | 6.27 | |
6) bimodal: pre-epidemic differs |
μ1A σA |
μ2A σA |
μ1B σA |
μ2B σA |
μ1B σA |
μ2B σA |
5 | -60.55 | 1.04 | |
Disruptive selection | 7) mixed: uni-bi.-bi., same modes |
μA σA |
μ1B σA |
μ2B σA |
μ1B σA |
μ2B σA |
4 | -60.63 | 1 | |
8*) mixed: uni.-bi.-bi., modes differ |
μA σA |
μ1B σA |
μ2B σA |
μ1C σA |
μ2C σA |
6 | -56.86 | 6.58 | ||
9*) mixed: uni.-bi.-uni. |
μA σA |
μ1B σB |
μ2B σB |
μC σC |
7 | -52.02 | 74 |
The mean/mode parameter of the unimodal normal distribution is μ and the standard deviation is σ. The central tendency parameters for the modes in the bimodal normal distributions are μ1 and μ2 and the standard deviation for each mode is σi. Within each model, parameters with the same subscript are constrained to be equal. Parameters for the best fit model (model 7, highlighted in bold) with 95% confidence intervals based on likelihood profile [51] in parentheses are μA = 0.28 (0.24–0.32), μ1B = 0.42 (0.38–0.46), μ2B = 0.10 (0.06–0.15), σA = 0.08 (0.07–0.11). Parameters for model 6 are μ1A = 0.21 (0.11–0.40), μ2A = 0.34 (0.11–0.40), μ1B = 0.42 (0.38–0.45), μ2B = 0.10 (0.06–0.14), σA = 0.07 (0.06–0.10). Model selection is based on Akaike's Information Criteria (AIC) which incorporates the number of estimated parameters (n). "Uni" = unimodal; "bi" = bimodal. The evidence ratio is the odds ratio of support for the best fit model over the listed model [50]; models with evidence ratios close to 1 are well-supported. Models that allow for a change in the distribution after sexual reproduction have an asterisk (*) after the model number.