Trapping experiments are a standard methodology in field studies of the chemical ecology of bark beetles and their associates. These experiments are typically designed to determine which semiochemicals are attractive compared to a blank control, as well as the relative attractiveness of different treatments, with differences among treatments being evaluated with multiple comparison procedures. One statistical problem that frequently arises is that the semiochemically blank control traps catch no insects and thus have zero variance. We show through a simulation study that this zero variance problem inflates both the per comparison and experimentwise error rates, through its effect on mean square error. The problem is especially severe for one commonly used method, Fisher's protected least significant difference, and the error rates are further inflated when other treatments in the experiment are unattractive. We also note that many recently published studies use multiple comparison procedures that are no longer recommended in the statistical literature. Possible solutions to these problems are presented.
Species 1: Coleoptera Scolytidae Dendroctonus frontalis (southern pine beetle)
Species 2: Coleoptera Cleridae Thanasimus dubius (checkered beetle)
Keywords: semiochemicals, traps
Back to Ten-Minute Papers, Section Cd. Behavior and Ecology
Back to Ten-Minute Papers, Section Ca, Cb, Cc, Cd, Ce, and Cf
Back to The 2003 ESA Annual Meeting and Exhibition