|Hurlbert, Stuart - SAN DIEGO STATE UNIV|
Submitted to: Journal of Economic Entomology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: July 1, 2002
Publication Date: July 1, 2003
Citation: HURLBERT, S.H., MEIKLE, W.G. PSEUDOREPLICATION, FUNGI, AND LOCUSTS. JOURNAL OF ECONOMIC ENTOMOLOGY. 2003. Interpretive Summary: 1. Problem. A paper was recently published, in the Journal of Economic Entomology, that reported results of a field experiment conducted over three weeks in November/ December, 1994, in Madagascar. The experiment involved treating 1 ha plots of uncultivated land with different strains of native entomopathogenic fungi. Each treatment was replicated twice (two 1 ha plots per treatment) plus a control were used, so the experiment was conducted over 10 ha. Two kinds of insect traps were used, and 3 of each trap type were placed on each plot (6 traps total). Traps were monitored 3 times after treatment - once every 10 days. For each treatment the authors pooled all data for different plots, different trap types and different sampling occasions, and they conducted nonparametric tests to determine if the density and diversity of the beetle fauna were affected by the fungal treatments. The authors did find that, according to their statistical tests, one isolate did apparently have an effect on beetle density and diversity, and they recommended that that isolate be tested further before release and that other researchers use their techniques. However, in all their statistical analyses the authors committed a major statistical error called pseudoreplication (first described by S.H. Hurlbert, who is senior author on this paper). Because the paper deals with the effects of entomopathogens on nontarget insects, it is important that the data are analyzed properly. 2. Approach. Our approach was very simple - we draw the author's attention to the error by describing the error, provide references concerning the seriousness of the error, and point out that none of the 142 statistical tests in the paper were valid and therefore none of the conclusions were valid. We suggest the proper statistical analysis, and we recommend the journal provide space for a corrected analysis by the original authors. 3. Results. The point of this paper is to draw the attention of the authors and journal readers to the seriousness and frequency of this type of error. That the statistics were improperly conducted does not mean that the fungi should not be tested further, nor that beetle density and diversity should not be taken into account in studies of this nature, but that the statistics used to arrive at those conclusions were invalid.
Technical Abstract: We review the statistical analyses and interpretations of an experiment on control methodologies for the migratory locust and their effects on beetle populations (Ivie et al., 2002, J. Econ. Entomol. 95:651), correct some of that paper¿s misinterpretations of the nature of pseudoreplication, point out that all 142 analyses presented represent a combination of temporal and sacrificial pseudoreplication, and determine that none of the paper¿s conclusions has any statistical support. We suggest a concise manner in which the core information from the study might be re-presented after reanalysis of data.