### Meeting Abstract

**68.5** Jan. 7 **Phylogeny, regression, and the allometry of physiological traits** *O'CONNOR, M.P.*; AGOSTA, S.J.; HANSEN, F.; KEMP, S.J.; SIEG, A.E.; MCNAIR, J.N.; DUNHAM, A.E.; Drexel University; University of Pennsylvania; University of Pennsylvania; University of Pennsylvania; Drexel University; Academy of Natural Sciences of Phildelphia; University of Pennsylvania* mike.oconnor@drexel.edu

Most functional allometric relationships are reported as ordinary least squares (OLS) regressions. Unfortunately such relationships usually involve independent and dependent variables, each of which includes intrinsic and measurement error for each data point (usually different species) and no clear causal mechanism designating one variable as causal and the other as a response variable. In addition, the data are, at least potentially, phylogenetically constrained rather than independent. None of these features are strictly consistent with the OLS regression model. Alternate models employing phylogenetic contrasts and major axis or reduced major axis (RMA) have been used but the extent to which they remedy the problems of OLS regression are often not evaluated. To explore the importance of regression methodologies and phylogenetic contrasts in estimating regression slopes for phylogenetically constrained data, we simulated Brownian diffusive evolution of functionally constrained characters. Both OLS and RMA regressions under and over-estimated regression slopes under different circumstances, but that a modified orthogonal (LSVOR) regression was less biased than either OLS or RMA regressions. The LSVOR technique is conceptually intermediate between OLS and RMA regression and depends on variance estimates likely to be available in allometric regressions but often unavailable for other analyses. For strongly phylogenetically structured data, failure to use phylogenetic contrasts as regression data resulted in overestimation of the strength of the regression relationship and a significant increase in the variance of the slope estimate.