• Version
• Download 6
• File Size 255.96 KB
• Create Date 1 January 1998
• Download

Statistical tests on data from black bear (Ursus americanus) research often have low power because of limited sample sizes and sometimes subtle effects. We used the assumption that den type (open nest, hollow tree, excavation, etc.) has some effect on first year cub survival to illustrate the use of statistical power in black bear research. We tested the hypotheses that den type does not affect minimum first-year cub survival (MFYS) in Massachusetts (MA) or Minnesota (MN), demonstrated the necessary sample sizes to conduct high power tests (≥0.80) of these hypotheses, and illustrated the use of power analysis in the design of bear research. Dens were assigned to 1 of 3 categories based on assumed thermal advantage. We used single factor analysis of variance to estimate effect of natal den type on MFYS with α ≤ 0.05 and to confidently conclude no effect with power ≥0.80. We obtained data on 47 litters in MA (1985-95) and 85 litters in MN (1982-94). For both states, we failed to reject the hypothesis that den type does not affect MFYS (MA: F = 0.63; 2,44 df; P = 0.539, power = 0.139; MN: F = 1.26; 2,82 df; P = 0.291, power = 0.258). However, the low power in each case precluded definitive conclusions regarding the effect of den type on MFYS. Achievement of power = 0.80, given the actual sample sizes in each case and α = 0.05, would have required large effect sizes. Given the observed effect sizes (MA = 0.155, MN = 0.166) and α = 0.05, total sample sizes of 395 litters in MA and 345 litters in MN would have been required to obtain power = 0.80. Our example illustrates the difficulty in testing hypotheses in black bear research. Although the MA and MN data represent 11 and 13 years of data collection respectively, neither generated sufficient sample sizes to adequately test a simple hypothesis with the design and analytical methods used in this study. Black bear researchers must consider power and draw only conclusions that are substantiated by their data.