Calculation of pairwise distances between population parameters
To test how genetic and phenotypic measures vary within island and mainland systems, we calculated the pairwise difference between population mean values for each phenotypic and genetic measure for each species 1) between island populations and 2) between populations on the mainland. To allow comparisons across the range of different measures, the pairwise difference between populations was expressed as the log response ratio of the paired measures, as follows:\(x_{\text{ij}}=\ \log_{10}(\frac{m_{\text{ij}}}{n_{\text{ij}}})\), where m = larger population mean value and n = smaller population mean value, i = population measure type , j = species. Higher log ratios indicate more variability between sites for phenotypic or genetic values (cf. Hedges et al. 1999). In this metric there was no directional structuring of the pairwise distances between populations, and values ranged between \(\log_{10}\)(1) and\(\log_{10}\)(∞). This dimensionless metric did not require further standardisation across different types of genetic and phenotypic variability measures, enabling meaningful comparisons between populations. The pairwise distance in genetic diversity between populations bears the following specificities: i) It quantifies difference in genetic diversity (diversity of genotypes in a population), not genetic differentiation (difference between genotypes) e.g., two strongly differentiated genotypes can have identical genetic diversity and thus, low value of the metric; ii ) It quantifies the scale of the difference, not the level of diversity e.g., comparing equally high diversity populations can be identical to comparing equally low diversity populations.
The non-independence of pairwise distance measures was accounted for in the statistical analysis (see below).