Table 2.
Statistical support for whether the distribution of phenotypes for eco-morphological traits is composed of one or two normal distributions.
| Character | Data | Goodness-of-fit | AICc for unimodal | AICc for bimodal | ΔAICc | Supported distribution |
|---|---|---|---|---|---|---|
| Lip size | Standardized area | P < 0.0001 | -139.83 | -448.46 | 308.64 | Bimodal (strong support) |
| Body shape | PC1 | P < 0.0001 | -486.00 | -1037.57 | 551.57 | Bimodal (strong support) |
| Pharyngeal jaw | PC1 | P < 0.0002 | -411.53 | -868.68 | 457.15 | Bimodal (strong support) |
| Diet | MDS 1 | P < 0.0001 | 83.53 | 64.05 | 19.48 | Bimodal (strong support) |
| Stable isotope (13C) | 13C defatted | P < 0.0008 | 139.41 | 134.52 | 4.89 | Unimodal = bimodal |
| Stable isotope (15N) | 15N | P < 0.0556 | 71.65 | 64.50 | 7.16 | Bimodal (moderate support) |
Data are composed of raw or multivariate data. Low P values for goodness-of-fit suggest poor fit of the data to a single normal distribution. ΔAICc is the difference in model AICc for a single- compared to a two-component model. Levels of support for one model over the other are described in the Methods. Frequency histograms and normality plots can be found in Additional File 7.
AICc, Akaike Information Criterion; MDS, multidimensional scaling.