predict.eicm {eicm}

Predict method for EICM fits

Description

Obtains probability predictions (conditional or unconditional) or a model realization from a parameterized EICM model.

Usage

## S3 method for class 'eicm'
predict(object, nrepetitions = 10000, newdata = NULL, ...)

Arguments

object

the fitted EICM model.

nrepetitions

the number of realizations to conduct for computing probabilities. Set to 1 if you only need simulated community data.

newdata

optionally, a matrix in which to look for variables with which to predict. If omitted, the original data (used to fit the model) is used.

...

not used.

Details

The interaction network of the model must be an acyclic graph. Predictions are obtained by realizing the model multiple times and averaging realizations, because there is not a closed-form expression for their calculation.

To obtain conditional predictions, include presence/absence columns with species names in newdata. Named columns for all the environmental predictors must always be included.

Value

A species x sample matrix with predictions. If nrepetitions=1, predictions correspond to one realization, otherwise they are probabilities.

Note

If the eicm was fit without regularization, unconditional predictions are numerically equal to those of simple binomial GLMs.

Examples

# Load the included parameterized model
data(truemodel)

# for reference, plot the species interaction network
plot(truemodel, type="network")

# Unconditional predictions
# let's fix environmental predictors at 0, for simplicity.
predict(truemodel, newdata=cbind(env01=0, env02=0))

# Conditional predictions
# predict probabilities for all species conditional on the
# known presence of sp011 (compare sp014 and sp004 with the above)
predict(truemodel, newdata=cbind(env01=0, env02=0, sp011=1))

# Propagation of indirect species effects
# predict probabilities for all species conditional on the known 
# absence (first line) and known presence (second line) of sp005 and sp023
predict(truemodel, newdata=cbind(env01=0, env02=0, sp012=c(0, 1), sp018=c(0, 1)), nrep=100000)

# Notice the effects on sp026 and the effect propagation to those
# species which depend on it (sp013, sp008)
# Also compare with unconditional predictions