predict: Predict species abundances in an ordination
Description
This function fits a Generalized Additive Model (GAM) for
each species in a data.frame against an ordination.
Usage
# S3 method for dsvord
predict(object,comm,minocc=5,dims=1:ncol(object$points),
family='nb',gamma=1,keep.models=FALSE,...)
Value
A list object with vector elements aic, dev.expl, adj.rsq, and matrix fitted.
Optionally, if keep.models is TRUE, a list with all of the GAM models fitted.
list element aic gives the model AICs for each species, dev.expl gives the deviance
explained, adj.rsq gives the adjusted r-Squared, and fitted gives the expected abundance
of each species in each sample unit.
Arguments
object
an object of class dsvord
comm
a community matrix or data.frame with samples as rows
and species as columns
minocc
the minimum number of occurrences to model a species
dims
which specific dimensions to include
family
the error distribution specifier for the GAM function;
can be 'nb' for negative binomial, 'poisson' for the
Poisson distribution, or 'binomial' for presence/absence data
gamma
the gamma parameter to control fitting GAM models
keep.models
a switch to control saving the individual GAM models
The predict function sequentially and independently fits a GAM model
of each species distribution as a function of ordination coordinates, using the
family and gamma specifiers supplied in the function call, or their defaults.
The function fits two or three dimensional models; if the length of dims is greater than
three the dimensions are truncated to the first three chosen.
See Also
calibrate for the complementary function that fits GAM models
for environment variables