Meeting Abstract

39.1  Jan. 6  Estimating organismal performance: maximum performance, repeatability, and phenotypic correlations between traits ADOLPH, S.C.*; HARDIN, J.S.; Harvey Mudd College ; Pomona College adolph@hmc.edu

Whole-animal performance is often a key link between physiology/morphology and individual fitness. Laboratory measurements of performance almost always exhibit within-individual variation (i.e., repeatability less than 1.0). Within-individual variation causes biased estimates of several performance parameters, including maximum performance and correlations between performance under different conditions (e.g., sprint speeds at different temperatures). The bias becomes more severe as the amount of within-individual variation increases. Statisticians have been aware of correlation bias (= attenuation) since Spearman�s work in 1904, but corrections for bias are rarely applied in integrative biology. We describe how knowledge of the within- and among-individual components of variance (or equivalently, repeatability) can be used to obtain an unbiased estimate of the correlation between two traits, such as performance under two conditions or between a performance trait and an underlying physiological trait. However, unbiased estimators have not yet been developed for maximum performance itself, or for correlations involving maximum performance. Estimates of maximum performance (as well as its repeatability) depend on the magnitude and pattern of intraindividual variability and the per-individual sample size. The empirical form of these relationships can be used for an approximate correction for bias. We illustrate these phenomena with data on burst sprint speeds of Sceloporus lizards. We also discuss how the allocation of sampling effort between the number of experimental subjects versus the number of measurements per individual influences the accuracy and precision of performance estimates.