Laserfiche WebLink
Estimates of allelic richness per locus and sample were calculated using rarefaction to <br />compensate for differences in sample sizes. A directional Wilcoxon Signed-Rank test <br />(Wilcoxon 1945) available through Vassar Stats: Web Site for Statistical Computation <br />was used to test for statistically significant differences (a = 0.05) in mean allelic richness <br />between samples, and adjusted for multiple tests with the sequentially rejective <br />Bonferroni method (Rice 1989). P-values for Fis per locus and sample were based on 900 <br />randomization of alleles within samples. Statistical significance was set at a = 0.05 and <br />adjusted for multiple tests using the standard Bonferroni method (Rice 1989). Global <br />values of F-statistics and variance components were estimated using the Weir and <br />Cockerham (1984) method; standard errors were derived by jackknifing over all loci. <br />To check the dataset for scoring errors and null alleles, the program Micro- <br />Checker (van Oosterhout et al. 2004) was used. Micro-Checker suggests the presence of <br />null alleles when there is a general excess of homozygotes for most allele size classes at <br />loci and samples. In the present study when the presence of null alleles was suggested, <br />the Brookfield 1 method (1996) was used to estimate null allele frequencies and adjust <br />the genotype frequencies in the dataset. This method provides a relatively conservative <br />estimate of null allele frequencies in cases where amplification was successful (at least <br />one band) for all DNA samples. <br />ARLEQUIN 3.0 (Excoffier et al. 2005) was used to calculate FsT values for <br />pairwise comparisons between samples. For each comparison, the P-value was derived <br />based on 1000 permutations of genotypes between samples to obtain the null distribution. <br />Statistical significance was set at a = 0.05 and adjusted for multiple tests using the <br />standard Bonferroni method (Rice 1989).