General program for determining instar distributions from headcapsule widths: example analysis of mountain pine beetle (Coleoptera: Scolytidae) data

Determining instar distributions from field-collected insect samples is a common problem in insect ecology. We describe a generalized computer program (H(CAP)) that determines an optimum separation rule based on the distribution of headcapsule widths. The program is initiated through visual interaction with the user to determine starting points for separation rules. H(CAP) then determines the optimum instar classification rules; estimates of means and standard deviations of headcapsule widths for each instar; estimated number in each instar; and probabilities of misclassification. Application of the program is illustrated by the analysis of 3 data sets of 10,000 observations of headcapsule widths obtained from field collections of larvae of the mountain pine beetle, Dendroctonus ponderosae Hopkins. This analysis resulted in finding significant differences between larvae collected from different hosts (ponderosa pine, Pinus ponderosae Lawson, and lodgepole pine, Pinus contorta Douglas variety latifolia Engelmann) and ecological circumstances (outbreak phase and geographic location). These results indicate that caution must be taken when extrapolating published results to new data or from one ecological situation to another