Add spatial only model to config file

[elizabethng]

For model selection, to see if spatial variation is warranted too (in the first component, that would be FieldConfig = c(1, 0, 0, 0)

[elizabethng]

Can also probably just do the gamma models, since those basically fit the best for everything.

Need

• st-full: FieldConfig = c(1, 1, 1, 1)
• st-pres: FieldConfig = c(1, 1, 0, 0)
• s0-full: FieldConfig = c(1, 0, 1, 0) skip this one for now
• s0-pres: FieldConfig = c(1, 0, 0, 0)
• st-none: FieldConfig = c(0, 0, 0, 0)

Model selection for additive mixed models (form Zuur)

1. Parameterize models with all fixed components we could possibly want ("beyond optimal")
2. Fit different random structures using REML
3. Check random effect variances and select the best random structure
4. Fit different fixed effects models using ML and compare
5. Re-fit the best model using REML

[elizabethng]

• [x] Make a fast_fit version of model with no plots, return just AIC and parameter values
• [x] Random effect values: epsilon is spatio-temporal, omega should be spatial
• [x] Run model selection for random effects with REML
• [x] Process output to check for top models - [ ] Run model selection for covariates using ML - [ ] Re-run top model using REML (and also bias correct?)

[elizabethng]

See other notes in https://github.com/elizabethng/predator-diets/issues/40

[elizabethng]

Added spatial only and non-spatial model in https://github.com/elizabethng/predator-diets/pull/50/commits/5cb2a6e092656e15a44632b5e050d1d69b7f4c1f