Ecological Archives E096-255-A4
Geneviève Lajoie and Mark Vellend. 2015. Understanding context dependence in the contribution of intraspecific variation to community trait–environment matching. Ecology 96:2912–2922. http://dx.doi.org/10.1890/15-0156.1
Appendix D. Statistical details on computing the contribution of environmental variables to SPT and ITV.
For each of the three CWMs and each of the three traits (nine combinations), we first conducted a multiple regression that included all environmental variables (both linear and quadratic terms), which we orthogonalized to minimize collinearity using the function poly in the {stats} package. We then performed stepwise procedures for each model to reduce the number of explanatory variables to be included in the analyses. For any given trait, all variables that remained in any of the three models were kept in the final set. For each component of trait variation (i.e., total, SPT and ITV), the contribution of a given environmental variable was calculated as the standardized regression coefficient in the reduced model, multiplied by the simple correlation coefficient of that variable with the CWM (Borcard 2002). Individual contributions of each variable sum up to the explained variation (R²) of the total model (Borcard 2002).
Literature cited
Borcard, D. 2002. Multiple and partial regression and correlation. Université de Montréal.