Pedigree metrics: What are they good for?

Joel L Pick , Barbara Class , Paul Acker, anne charmantier, Niels J. Dingemanse, Ned Dochtermann, Julien G. A. Martin, Denis Réale, Jane Reid, Holger Schielzeth, Céline Teplitsky, Kasha Strickland, Yimen Gerardo Araya Ajoy

Estimating additive genetic variance (V_A) is key to understanding evolutionary potential in wild populations. V_A is commonly estimated using relatedness data from pedigrees in so-called `animal models'. The accuracy of these estimates depends strongly on information in the pedigree. Typically metrics describing pedigree structure (such as the number of individuals, maternities, pedigree depth etc.) are provided alongside V_A estimates. However, what these metrics represent is not well understood. Here, we first explored the evolution of metric reporting. Then, using 53 wild pedigrees, we demonstrated that precision of heritability (h^2; ratio of V_A to total phenotypic variance) estimates was strongly predicted by pedigree size, but other metrics provided little additional information about precision. An seldom-used analytical approximation almost exactly reproduced simulated standard errors for h^2 estimates from pedigrees and genomic relatedness matrices, offering an alternative to simulation-based power analysis for simple models. Although the utility of commonly reported metrics is limited for precision, they provide important information about risk of bias from confounding effects, which is valuable for planning future studies and assessing published ones. We demonstrated this using the example of unmodelled maternal effects. Future work on biases in h^2 estimation should explicitly take pedigree structure into account.