Snakemake workflow for running FEEMS, and comparing it to other methods (IBD-FEEMS which is FEEMS null model, Locator a neural network based approach) Attempted to also include SPASIBA in the pipeline but could not get it working.
Be sure to use Karl's fork of FEEMS, which includes spatial prediction functions: https://github.com/karltayeb/feems
Check out notebooks/peter2020.ipynb to see the comparative performance of FEEMS, IBD-FEEMS, and Locator on Peter 2020 Data
Run notebooks/peter2020_assignment_uncertainty.ipynb for interactive plots to explore uncertainty in FEEMS assignments
Check out notebooks/wolves.ipynb to see the comparative performance of FEEMS, IBD-FEEMS, and Locator on wolf migration example
Run notebooks/wolves_assignment_uncertainty.ipynb for interactive plots to explore uncertainty in FEEMS assignments
Trying to identify scenarios where FEEMS clearly outperforms IBD-FEEMS. You can start at notebooks/feems_vs_ibd_anisotropic_sim.ipynb. In this simulation both models work comparably, but you can use this as a starting point for making new simulations that challenge one/both methods.