Growing the biphasic framework: Techniques and recommendations for fitting emerging growth models

Several new growth models have been proposed to account for the life-history trade-offs that occur when indeterminately growing species allocate energy between somatic growth and reproduction. These models can improve the understanding of lifetime growth and life history, but can be more difficult to fit than conventional growth models. Increased data demands, multiple growth phases and increased parameterization all serve as barriers to the adoption and proper use of these new models. We review and comment on confounding issues during model fitting for several of these models, and provide advice on surmounting such issues. We then simulation-test an example model, the Lester biphasic growth model, using several common fitting approaches. We highlight the biases and precision of each approach and provide guiding documents using r and jags code. The Bayesian Markov chain Monte Carlo and likelihood profiling approaches generally provided the best fits. Simpler approaches can be unbiased and precise if sampled data are of relatively high quality (e.g. moderate sample sizes for juvenile and adult phases) and model assumptions are met. Bayesian hierarchical approaches can accommodate more complicated data scenarios (e.g. unbalanced design across multiple populations); we provide an example of such an approach by recovering growth trajectories and inferring growth-associated trait variation and environmental effects across multiple populations. Conventional growth models provide limited inference on life history. Many biphasic growth models can provide direct inference on multiple life-history traits, but can be difficult to fit. The recommended approaches herein provide a path forward for fitting biphasic growth models in a variety of scenarios, allowing for wider application and tests of life history and ecological theory.