Context dependence of biodiversity and ES recovery
We used linear mixed-effects models to evaluate whether the effects of restoration on biodiversity and ES varied with context. Context was parameterized using four nominal fixed factors (ecosystem type, main cause of degradation, restoration action, and experimental design) and the continuous fixed factor of restoration age, defined as the decimal logarithm of the number of months between completion of the last restoration action and evaluation. We added a fifth nominal fixed factor with two levels (biodiversity or ES) because we used RRs for both biodiversity and ES recovery in the analysis. Study site was the random-effect factor and RR was the dependent variable.
We also built a second model in which we reduced the degrees of freedom by including only factor categories containing at least 30 observations. Since this reduced the average sample size in each category, we discarded this model in favor of the first. Finally, we applied a backward elimination procedure in which non-significant terms (p