Models of leaf morphology as a function of habitat and altitude.

Note that all models fitted with interaction terms always include the same variables as main effects. (a) The fits of alternative models are compared by ΔAICc and Akaike weights averaged across 1000 trees. (b) Parameter estimates are the median estimates from pgls models fitted to 1000 trees. Leaves per frond: two models including habitat but no interaction term are essentially indistinguishable based on AICc, we present the results from the more complex model (Habitat + Altitude). The median maximum likelihood estimate of Pagel's lambda across trees for this model is 0 and is only non-zero in analyses with 18 from 1000 trees. Consequently, the model is equivalent to one in which phylogeny is not included, and hence there is no variation in parameter estimates attributable to variation in phylogeny. The adjusted R2 for the model is 0.901. Pinnae length: the best-fitting model including an altitude x habitat interaction term explains 56.8% of the among-species variation in pinnae length (95% quantiles for adjusted R2 = 0.513–0.735). The median estimate of Pagel's lambda for the interaction model across 1000 trees is 0.779 (95% quantiles = 0.000–0.957) and greater than 0.5 in analyses on 872 trees.
Models of leaf morphology as a function of habitat and altitude.