In this paper we re-analyze previously published data regarding the response of several prey populations to manipulation of predaceous larval dragonfly (Insecta: Odonata) densities in four separate field enclosure experiments. Using a computer-intensive "rerandomization" approach to testing hypotheses, we show that the individual experiments were not sufficiently powerful to consistently reject false null hypotheses. Combining the data from three comparable experiments, we can enhance the power associated with such tests. Three prey categories (Trichoptera, Oligochaeta, and large Cladocera), constituting less than one-third of the typical odonate diet, were found to be consistently depleted in enclosures with odonate larvae; but the extent of their depletion was not increased at high (ambient) compared with low (half-ambient) odonate densities. These results support our previously published conclusions that exploitation competition was not an important phenomenon for odonate larvae in these experiments.