Full Subsets Multiple Regression Using GAMs

Full-subsets information-theoretic approaches are increasingly used to explore predictive power and variable importance when a wide range of candidate predictors are being considered. This package provides functions that can be used to construct, fit, and compare a complete model set of possible ecological or environmental predictors for a given response variable of interest. Models are based on Generalized Additive Models (GAMs) and build on the 'MuMIn' package. Advantages include the capacity to fit more predictors than there are replicates, automatic removal of models with correlated predictors, and support for model sets that include interactions between factors and smooth predictors, as well as smooth-by-smooth interactions via te(). Methods are described in Fisher et al. (2018) .

FSSgam

FSSgam constructs, fits and compares a complete model set of candidate ecological or environmental predictors for a response variable of interest. Models are generalized additive models, fitted with mgcv or gamm4, and ranked by AICc using MuMIn.

The approach admits more predictors than there are replicates, removes models with correlated predictors automatically, and supports model sets containing interactions between factors and smooth predictors as well as smooth-by-smooth interactions through te(). The method is described in Fisher et al. (2018), Ecology and Evolution, doi:10.1002/ece3.4134.

Installation

# install.packages("devtools")
devtools::install_github("beckyfisher/FSSgam_package")

Usage

An analysis is built in two steps. generate_model_set() builds the candidate set from a test fit and a list of predictors, and fit_model_set() fits and compares that set. Separating them allows the candidate set to be inspected before anything is fitted, and allows the fitted objects to be discarded as they are summarised, which matters for large sets.

library(FSSgam)
library(mgcv)

data(case_study1)
use.dat <- case_study1
use.dat$site <- as.factor(use.dat$site)

test.fit <- gam(Herbivore.abundance ~ s(depth, k = 3, bs = "cr") + s(site, bs = "re"),
                family = tw(), data = use.dat)

model.set <- generate_model_set(
  use.dat        = use.dat,
  test.fit       = test.fit,
  pred.vars.cont = c("complexity", "depth"),
  pred.vars.fact = "ZONE",
  null.terms     = "s(site,bs='re')",
  max.predictors = 2,
  k              = 3
)

out <- fit_model_set(model.set)
out$mod.data.out          # the model table, ranked by AICc
out$variable.importance   # summed model weights per predictor, by criterion

full_subsets_gam() performs both steps in one call. It saves every fitted model, so it is recommended only for small candidate sets.

check_correlations() and check_non_linear_correlations() report the predictor correlations the model set is screened against, and can be called before generating a set.

Documentation

Function reference: https://beckyfisher.github.io/FSSgam_package/

Worked case studies, an FAQ and the material accompanying the publication are maintained in the companion repository, which is the citable reference for the method: https://github.com/beckyfisher/FSSgam and https://beckyfisher.github.io/FSSgam/

Citation

Fisher R, Wilson SK, Sin TM, Lee AC, Langlois TJ (2018) A simple function for full-subsets multiple regression in ecology with R. Ecology and Evolution 8(12): 6104-6113. doi:10.1002/ece3.4134

License

The code is released under the Apache License 2.0

Copyright 2020 Australian Institute of Marine Science

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at 

   http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.