Estimates, bootstrapped confidence intervals and likelihood ratio test statistics of the reduced optimal models (see Methods for details) for multivariate linear (β) and quadratic (γ) selection gradients at ridge and snowbed sites, using a proxy for sexual reproductive fitness (flowering probability) and a proxy for clonal reproductive fitness (relative stem ratio between 2013 and 2012).

Note that a quadratic term was kept only in the reduced optimal model for the analysis using clonal fitness at snowbed sites. We also report delta AIC (dAIC) values for the comparisons between the model from which the respective parameter has been removed and the reduced optimal model. Estimates, bootstrapped confidence intervals and likelihood ratio test statistics of the reduced optimal models (see Methods for details) for multivariate linear (β) and quadratic (γ) selection gradients at ridge and snowbed sites, using a proxy for sexual reproductive fitness (flowering probability) and a proxy for clonal reproductive fitness (relative stem ratio between 2013 and 2012).