Ecological Archives A022-023-A2

Richard Bischof and Jon E. Swenson. 2012. Linking noninvasive genetic sampling and traditional monitoring to aid management of a transborder carnivore population. Ecological Applications 22:361–373.

Appendix B. Cross-validation.

With the validation data (see main text), we conducted tests to answer the following questions:

  1. How does the distribution of the 95% kernel home range size estimate for the cross-validation data set (including only females that produced cubs) compare with that of the set of home ranges used during simulations in the model?
    Result: The distributions of the 95% kernel home range sizes of the cross-validation set (females with cubs, N = 13) and the simulation set (N = 35) show substantial overlap. No significant difference between their means was detected (t = 1.49, df = 16.62, P = 0.155, Fig. B1).
  2. How does the actual proportion of utilization distributions (UD) attributable to the southern bear region in Sweden (for reproducing females) compare with the model-simulated proportion for the same region?
    Result: The predicted proportion of annual UDs for cross-validation females that falls within the southern bear region in Sweden (81.3%) corresponded closely with the proportion predicted through using the home range simulation submodel of the model (84.8%, Fig. B1).
  3. How does the actual number of reproductions in the cross-validation dataset compare with the predicted number of reproductions using the logistic regression model described in the main document?
    Result: The number of observed reproductions (11.6 reproductions for 91 bear-years) and the number of reproductions (13) predicted using the logistic regression model (applied to the known age structure) were comparable (Fig. B1).

FIG. B1. Cross-validation results: left panel - a comparison of the distribution of log-transformed 95% kernel home range sizes of the cross-validation data set (red line, mean: red triangle) and the home range sample used during simulations (black line, mean: black triangle); middle panel - the distribution of the model-predicted proportion of utilization distributions within a focal area (southern bear region in Sweden) and the value of the actual proportion of utilization by cross-validation bears-years within the same area (red triangle); right panel distribution of the GLM-predicted number of reproductions (black line, mean: black triangle) and the number of actual reproductions during cross-validation bear-years (red triangle). 95% confidence limits for the distributions are marked by dashed vertical lines.

[Back to A022-023]