The underlying sources of growth variability in a population cannot generally be known, so when modelling growth it is important to understand the consequences of assuming an incorrect error structure. In this study, four error models for a von Bertalanffy growth curve with asymptotic length parameter L∞ and growth rate parameter k are considered. Simulations are carried out in which data are generated according to one of the models and fitted assuming each of the models to be true. This is done for two types of data: direct age–length and tag–recapture. For direct age–length data, the consequences of not accounting for individual growth variability, or assuming the wrong source of variability, are minor, even when individual variability is high or data coverage is poor. For tag–recapture data, some substantial biases in growth estimates can arise when individual variability exists but is not accounted for. Importantly, however, incorporating variability in just one parameter (be it L∞ or k), even if the variability truly stems from the other or both parameters, generally leads to much smaller biases than assuming no individual variability. Often the alternative models cannot be distinguished using standard model selection procedures, so caution is warranted in using model selection to draw inferences about underlying sources of growth variability.
Correlated Response on Growth Traits and Their Variabilities to Selection for Ovulation Rate in Rabbits Using Genetic Trends and a Cryopreserved Control Population
