2.4 Estimation of sex-biased dispersal using nuclear markers
To detect sex-biased dispersal, we separated sequences from females and males into two separate datasets, excluding three populations (Saint-Barthélémy, Funchal and Selvagem) represented by fewer than five individuals from each sex. Sex-biased dispersal was tested at both intra- and inter-lineage scales, expecting females to be less structured than males since males are presumably more philopatric (Greenwood 1980). We calculated average pairwise relatedness for each sex, within each population, using the triadic likelihood estimator (Wang 2007) implemented in Coancestry (Wang 2011). To test if the difference of mean relatedness between males and females of each population was significant, we used the test of difference between sex by bootstrapping samples 1000 times and recalculating difference in means between sex for each bootstrap. Observed and simulated differences was then compared, and if the observed difference fell outside of the 95% confidence interval, we considered it to be significant.
If females disperse more than males, females sampled from a single population will be a mixture of residents and immigrants. The female sample will therefore deviate from the Hardy-Weinberg equilibrium and show a deficit of heterozygotes (Wahlund effect).F IS calculated for the female sample is thus expected to be larger than the male F IS (Goudet et al. 2002). We estimated F IS separately for females and males for all tested populations, and evaluated its significance using 1000 permutations with Genetix. Conversely, we expectF ST (Goudet et al. 2002) to be higher in philopatric males than in females. We calculatedF ST for each pair of populations within the two datasets, with Arlequin.