highly significantp < 0.001
Why do site-based and species-based analyses differ with these data? An explanation is that there is an ‘unobserved’ trait, namely Specific Leaf Area (SLA), that is almost uncorrelated with Spread (correlation 0.02) and that strongly interacts with Snow as both the site-based and species-based tests are highly significant using either the fourth-corner correlation or the GLM tests using Poisson log-likelihood (all p < 0.001, yielding p max < 0.001).