Measuring standardized effect size improves interpretation of biomonitoring studies and facilitates meta-analysis

Intersite differences in benthic communities detected by tests of null hypotheses are used routinely to infer effects of habitat perturbation. The statistical outputs from these tests are often treated as binary results (presence or absence of detectable effects), and the sizes and potential biological importance of detected differences or effects are frequently ignored. This situation can be remedied by measuring standardized effect sizes of detected differences. To demonstrate the benefits of standardized effect sizes, we compared benthic communities in streams draining forested and perturbed catchments based on kick-net samples, and samples from bricks and Hester–Dendy multiplate samplers. We complemented null hypothesis testing by calculating standardized effect sizes (Cohen's d) and their confidence intervals (CIs) to rank 14 benthic metrics and the 3 sampling techniques. Despite having higher variance than metrics from brick or Hester–Dendy samplers, metrics from kick-net samples better separated sites than did metrics from bricks or Hester–Dendy samples. Metrics from brick samples separated sites more often and by a larger number of standard deviations than did metrics from Hester–Dendy samplers. Metrics that included mayfly abundance or richness produced the largest d-values, particularly when calculated from kick-net samples. Metric rankings were inconsistent among techniques. Successional changes over the 30-d study were subtle or absent, but generally consistent among sampling techniques. Differences detected with few replicate kick-net samples were consistent in direction but smaller than differences detected with more replicates. In some cases, differences with large d-values were not detectable with small sample sizes and standard null-hypothesis-testing approaches. These differences were consistently confirmed by addition of replicates. d-values and their CIs add value to data sets, particularly given the small number of replicates common in labor-intensive ecological studies. This approach expresses biological differences in a common currency that can be compared across studies regardless of units of measurement, scale, or technique.