Ecological Archives C006-069-A3

J. D. DeVoe, R. A. Garrott, J. J. Rotella, S. R. Challender, P. J. White, M. O'Reilly, and C. J. Butler. 2015. Summer range occupancy modeling of non-native mountain goats in the greater Yellowstone area. Ecosphere 6:217. http://dx.doi.org/10.1890/es15-00273.1

Appendix C. Model selection results.

Table C1. AIC model selection results arranged in model suites used for obtaining a complex model to estimate overdispersion (ĉ) for examining relationships of habitat covariates and occupancy (ψ) probability of sampling units for mountain goats in the greater Yellowstone area, USA.

TableC1

Notes: The top models within 5 ΔAIC units of the overall top-ranked model within each suite are presented. Covariate subscripts indicate spatial scale and superscripts indicate functional form (2=quadratic, ps=pseudothreshold, [none]=linear). Detection (p) probability was parameterized using the 100-m scale pseudothreshold form of COV. K = no. of parameters. Negative log likelihood for the top model from the combined suite was 1932.87.


Table C2. Estimated coefficients and 95% confidence intervals for the top competitive models from combinations of top models from the four model suites (sloped terrain, ruggedness, forage, and heat load) based on QAIC model selection, examining predicted effects of covariates on mountain goat occupancy (ψ) and detection (p) probabilities of sampling units in the greater Yellowstone area, USA.

TableC2

Notes: Superscripts are as in Table 2. ID number for reference to results in Table 3.

 

Table C3. Post hoc QAIC model selection for examining relationships of covariates on detection (p) probability of mountain goats for sampling units in the greater Yellowstone area, USA.

TableC3

Notes: Occupancy (ψ) probability was parameterized in each model based on the inferential model from the combined QAIC model suite results. Covariates on detection (p) probability were measured at the 100-m spatial scale and are included in parentheses, with superscripts as in Table 2. Negative log likelihood for the top model was -848.90.


[Back to C006-069]