(1) Parentage analyses
Parentage was assigned using Cervus 3.0 (Kalinowski, Taper, & Marshall, 2007) based on autosomal loci. Cervus compares likelihood ratios of two most likely candidate parents and assigns parentage based on statistical thresholds generated during the simulation analysis. For Cervus analysis, we used a simulation of 100,000 offspring, an error rate of 0.01, 90% relaxed and 95% strict confidence level, and accounted for relatedness of candidate mothers and fathers. Relatedness was calculated with the R package related 1.0 (Pew, Muir, Wang, & Frasier, 2015) using Wang’s estimator r (Wang, 2002). This estimator was chosen because it performed best in simulations, showing the highest correlation between observed and expected values for our set of loci. Additionally, we used Colony 2.0.6.5 (Jones & Wang, 2010) to verify parentage assignments from Cervus. Unlike Cervus, Colony reconstructs a full pedigree, inferring sibship and parentage among individuals by comparing the likelihood of different clusters of individuals and maximizing group rather than pairwise likelihoods. For this analysis, we used an error rate of 0.01, male and female polygyny, and a sibship size prior of 1.6, calculated as the average number of young per family group in our study population. For both Cervus and Colony analyses, the set of candidate fathers included all sampled adult males plus the oldest subadult male from Group 6 that had dispersed from his natal group in the beginning of the study and could have sired offspring by the end of the sampling period. The set of candidate mothers included all adult females that shared their mtDNA haplotype with candidate offspring. For seven offspring (Supplementary Table 1), the mothers were known because they were seen nursing them. To test the reliability of our parentage estimates, we ran the analyses twice, with and without the respective set of known mother-offspring pairs. Combined non-exclusion probability for the set of 16 autosomal loci (with chr01b excluded) was 9.9x10-5 for the first parent, 3.4x10-7 for the second parent, and 9.0x10-12 for the parent pair, calculated using Cervus.
Additionally, we used the data for the X-linked locus chrXa to check for allelic mismatches between candidate parents and offspring. This locus was treated separately because neither Cervus nor Colony can handle a mixture of autosomal and sex-linked loci.