Recommendations for treated-area choice assays with termites (Isoptera)
The repellency of catnip (Nepeta cataria) essential oil was evaluated in a treated-area choice assay with subterranean termites (Reticulitermes spp.). It appeared that fewer R. virginicus were found on the treated portion of a petri dish within a period of about 7d; R. flavipes was not affected by the presence of the oil. The data collected from the control dishes, however, showed an unacceptably high Type I error rate (rejection of H0:nt=nu, when Ho is true, where nt is the expected number of termites on the treated side and nu is the expected number of insects on the untreated side). The tendency of termites to cluster was the probably reason for this, and calls into question the data obtained from the tests. Computer simulations, using a range of cluster factors, replications and numbers of termites per replication, were conducted, and the Type I error rate was calculated when a cluster center (a point corresponding to a random angle O and radius p value withing a circle) and random "termite" positions about the cluster center were plotted. Type I error rates were consistently inflated for all analyses that were based on individual animal behavior, and could not be corrected by increasing replications. We recommend an analysis where the number of insects on the treated or untreated side is recorded, then the dish is designated as "repelled" or "not repelled," and the number of repelled dishes is analyzed. This method effectively controls the Type I error rate so that it is no greater than the nominal value. It is recommended to use at least 25 replications to ensure adequate statistical power. Analysis or Boschloo's Exact Unconditional test.