• Version
• Download 10
• File Size 328.40 KB
• Create Date 1 January 1987
• Download

We used simulated grizzly bear (Ursus arctos) harvest data to answer 2 questions: (1) can we use 2-group discriminant function analysis to distinguish harvest age structures from populations that have equilibrated from those beginning to decline from overharvest? and (2) how powerfully can it distinguish with small samples typical of grizzly bear data? Simulated populations were subjected to experimental harvests to determine the harvest level that caused chronic declines. For each run, statistics that described age structures were calculated. We constructed a linear discriminant function equation based on the descriptive statistics that separated declining from equilibrating populations, and estimated the power of the equation using decline as the null hypothesis. Accepting a 10% chance of failing to detect a decline (Type I error), the equation had little power to correctly classify equilibrated populations. With sample sizes from a 3-year period in the range 72-153 animals, estimated power was about 50%. With 3-year sample sizes in the range 24-51, power was roughly 20%-25%. Only when comparing severely overharvested populations with equilibrated populations could power be raised to >60%. We conclude that detecting grizzly bear population declines at their outset will be unreliable when based solely on harvest age structure data.