Meeting Abstract
Quantifying the patterns and strength of phenotypic selection in the wild is fundamental to explaining causes of adaptive evolution. The advent of a theoretical framework to distinguish the direct and indirect components of multivariate selection on correlated phenotypic traits catapulted the study of selection, with thousands of estimates now published. Syntheses have yielded powerful insights into phenotypic selection, including that selection tends to be weak and that stabilizing selection was as common as disruptive selection. Although these patterns could be due to biological and environmental determinants, there is growing recognition that variation in computational methods is also important factors. A synthesis of regression-based methods for quantifying phenotypic selection was conducted via data aggregation, in which the raw data were re-interpreted using a systematic workflow to directly compare the effects of specific procedures on patterns of selection. Analyses were conducted on several empirical and simulated longitudinal datasets on morphology, and a sensitivity analysis was conducted on each dataset to evaluate how different model parameters, standardizations, regression types, inclusions of nonlinear selection, etc. impact selection coefficients. Preliminary results suggest that variation in the workflow generally did not alter which traits were under direct selection, but did change the strength of selection. An R package is in development to provide open-source datasets and computer code for reproducible evaluations of phenotypic selection to facilitate future syntheses. Evolutionary biologists are now equipped with an arsenal of tools to quantify and visualize the multivariate nature of selection, and a synthesis on these theoretical tools provide new perspectives on the patterns and strength of selection.
