Statistical analysis
We compared the NCE of fish on zooplankton densities across the four experiments for the subset of zooplankton taxa that were represented in all experiments (see Results, Table 1). We used a multivariate approach to evaluate the overall response of the zooplankton community to the presence of fish kairomones, and our initial analysis simultaneously tested the effect of experiment, treatment and the interaction between experiment and treatment on the subset of taxa using a permutational multivariate analysis of variance (PERMANOVA, McArdle and Anderson 2001). This analysis uses dissimilarity matrices, and we chose to use a dissimilarity matrix based on Kulczynski distances, which are particularly robust in identifying patterns in community data (Faith et al. 1987). Probabilities from the analysis were calculated based on 9,999 permutations. These effects were also investigated graphically by comparing experiments and treatments across different axes generated through non-metric multidimensional scaling (NMDS). The NMDS was based on Kulczynski distances and was constrained to three dimensions, which was the minimum number of dimensions that resulted in a stress measure less than 0.20 (Clarke 1993).
The effect of treatment on the density of individual taxa within an experiment was analyzed using quasi-Poisson regressions.