Gaussian Process regression for Empirical Dynamic Modeling

Functions for fiting hierarchical, separable length scale GP models with automatic relevance determination (ARD) for use in Empirical Dynamic Modeling (EDM) and other applications.

Details

The main function is fitGP which is used to train the model and can also produce predictions if desired. Use predict.GP to generate other or additional predictions from a fitted model, plot.GPpred to plot observed and predicted values and getconditionals to obtain condictional reponses. Also available for use is the function makelags which can be used to create delay vectors. See the (yet to be created) vignette for more detailed instructions.

Author

Original MATLAB code by Stephan Munch, adapted to R by Tanya Rogers.

Maintainer: Tanya Rogers <tanya.rogers@noaa.gov>

We thank Bethany Johnson, Lucas Medeiros, and Vadim Karatayev for beta testing and improvement suggestions.

References

Munch, S. B., Poynor, V., and Arriaza, J. L. 2017. Circumventing structural uncertainty: a Bayesian perspective on nonlinear forecasting for ecology. Ecological Complexity, 32:134.

Johnson, B., and Munch, S. B. 2022. An empirical dynamic modeling framework for missing or irregular samples. Ecological Modelling, 468:109948.