eicm (version 1.0.3)

generateEICM: Generate EICM model following a Beta frequency distribution

Description

Generates a randomly parameterized EICM model (predictors, environmental coefficients and species interactions), ensuring that communities simulated with this model have a frequency distribution that matches the given Beta distribution of frequencies as much as possible.

Usage

generateEICM(
  nspecies,
  nsamples,
  nenv,
  ninteractions,
  shape1,
  shape2,
  nbins = 10,
  nrepetitions = 5,
  shrinkage = 2,
  bounds = 10,
  swarm.size = floor((ninteractions + nspecies * (nenv + 1)) * 0.5),
  maxit.stagnate = 150
)

Value

A EICM model of class eicm

Arguments

nspecies

the number of species to generate.

nsamples

the number of samples to generate.

nenv

the number of environmental predictors to generate.

ninteractions

the number of species interactions to generate.

shape1

the shape1 parameter of the Beta distribution.

shape2

the shape2 parameter of the Beta distribution.

nbins

the number of histogram bins for comparing distributions.

nrepetitions

the number of times to realize a model candidate to average their distributions.

shrinkage

the shrinkage factor for generated coefficients, when computing fitness criterion. Ensures that the generated coefficients remain at plausible values.

bounds

the allowed range for the coefficients c(-bounds, +bounds).

swarm.size

the swarm size of the particle swarm optimization.

maxit.stagnate

the number of iterations without improvement until the optimization stops.

Details

This function is useful for generating a realistic random model for simulation, i.e. a model that, when simulated, will yield species communities with a distribution of frequencies akin of real communities: with most species being rare. The generated coefficients are not assumed to follow any distribution, but are optionally shrinked so that their values will remain "decent". The values of the environmental predictors are drawn from a gaussian distribution.

Examples

# \donttest{
# Generate model with 32 species, 30 species interactions and 2 environmental predictors
# for 500 samples with a frequency distribution following a Beta(1.5, 3)
model <- generateEICM(nspecies=32, nsamples=500, nenv=2, ninteractions=30,
  shape1=1.5, shape2=3)

# make one realization
data <- predict(model, nrepetitions=1)

# plot frequency histogram: should follow a Beta distribution.
hist(apply(data, 2, sum) / nrow(data), breaks=seq(0, 1, by=0.1), xlim=c(0, 1),
  main="Frequency distribution of one realization", xlab="Frequency in samples",
  ylab="Number of species")
# }