optimizePenalty-methods.RdIdentify the optimal value of the penalty term for unmarked models that
support penalized likelihood. For each potential value of the penalty term,
K-fold cross validation is performed. Log-likelihoods for the test data in
each fold are calculated and summed. The penalty term that maximizes the sum
of the fold log-likelihoods is selected as the optimal value. Finally, the
model is re-fit with the full dataset using the selected penalty term.
Right now only Bayes-inspired penalty of Hutchinson et al. (2015) is supported.
Currently the only fitting function that supports optimizePenalty is
occuMulti for multispecies occupancy modeling; see Clipp et al. (2021).
# S4 method for unmarkedFitOccuMulti optimizePenalty( object, penalties = c(0, 2^seq(-4, 4)), k = 5, boot = 30, ...)
| object | A fitted model inheriting class |
|---|---|
| penalties | Vector of possible penalty values, all of which must be >= 0 |
| k | Number of folds to use for k-fold cross validation |
| boot | Number of bootstrap samples to use to generate the variance-covariance matrix for the final model. |
| ... | Other arguments, currently ignored |
unmarkedFit object of same type as input, with the optimal
penalty value applied.
Clipp, H. L., Evans, A., Kessinger, B. E., Kellner, K. F., and C. T. Rota. 2021. A penalized likelihood for multi-species occupancy models improves predictions of species interactions. Ecology.
Hutchinson, R. A., J. V. Valente, S. C. Emerson, M. G. Betts, and T. G. Dietterich. 2015. Penalized Likelihood Methods Improve Parameter Estimates in Occupancy Models. Methods in Ecology and Evolution. DOI: 10.1111/2041-210X.12368
Ken Kellner contact@kenkellner.com