2.4. Population genetic diversity
We used the populations module in STACKS v.2.61 (Catchen et al., 2013) to calculate diversity statistics per population, including the observed heterozygosity (Ho), expected heterozygosity (He), nucleotide diversity (Pi), inbreeding coefficient (Fis) and number of private alleles (PAs). Allelic richness (Ar) was calculated with the ‘hierfstat’ package in R (Goudet et al., 2015) using four random individuals per population, i.e. the minimum number of samples per population (minimum number of alleles, min.all=8). Significant associations between genetic diversity indexes (He, Pi, Ar, Fis) and geographic variables (islet surface area, maximum altitude, and the shoreline development index, SDI, i.e., a measure of island perimeter complexity, all log-transformed) were assessed with simple and multiple linear regression in R.