Comparing the performance of analytical techniques for genetic PARENTAGE of half-sib progeny arrays

The prevalence of female multiple mating in natural populations is important for many questions in mating system evolution. Several statistical techniques use genetic data to estimate the number of fathers that contribute gametes to broods, but they have not been widely compared to assess the magnitude of differences in their performance. With a combination of new data and reanalysis of previously published data, we compared five analytical approaches: (1) allele-counting, (2) parental reconstruction in GERUD, (3) a Bayesian probability model to estimate the frequency of multiple mating (FMM), (4) computer simulations based on population allele frequencies in HAPLOTYPES and (5) Bayesian parental reconstruction in PARENTAGE. The results show that choice of analysis technique can significantly affect estimates of sire number. Estimates from GERUD conformed exactly to results obtained from strict exclusion of potential sires in an experimental context. However, estimates yielded by HAPLOTYPES and PARENTAGE sometimes exceeded the numbers from GERUD by as much as 120 and 55%, respectively. We recommend GERUD over these other approaches for most purposes because of its accuracy and consistency in this analysis. Our novel genetic data set allowed us to investigate the extent and frequency of multiple paternity in a marbled salamander (Ambystoma opacum) population in South Carolina, USA. A. opacum contrasted with other salamander species by having relatively low levels of multiple paternity (only 31-54% compared with 71-96%). Although A. opacum had the lowest level of multiple paternity under all analytical approaches used here, the magnitude of differences among species varied.