RAD (version 0.3)

MDMnMod: MDMnMod

Description

MDMnMod

Usage

MDMnMod(MDMn.form, data, ID, dist="MDMn", scale.covar=FALSE, est.var=TRUE, calc.resid=TRUE, trace=TRUE)

Arguments

MDMn.form

an object of class "formula" (or one that can be coerced to that class):a symbolic description of the model to be fitted. Must include both the response and covariates

data

a data frame containing the variables in the formula. Must contain a column with the ordered (most to least) abundances for each site. Covariates must be repeated for each row belonging to a site.

ID

Vector of site identifiers for all observations. For every abundance (row) that comes from a site, the same identifier must be used. The length of ID must equal nrows(data)

dist

one of either multinomial "Mn" or Dirichlet Multinomial "DMN" or Modified Dirichlet Multinomial "MDMn" (default)

scale.covar

Should the model matrix be scaled (TRUE/FALSE)? Useful if models do not converge. Setting this to TRUE will save the column means and sd for later prediction.

est.var

estimate the variance-covariance matrix using newton-raphson

calc.resid

should residuals be calculated?

trace

print model trace

Value

coef

coefficents

vcov

variance-covariance matrix

logl

log-likelihood

AIC

AIC

residuals

model residuals, calculated using PIT

fitted

matrix of fitted values (nij and pij)

mean.X

column means for model matrix if scale.covar is TRUE

sd.X

column sds for model matrix if scale.covar is TRUE

formula

model formula

Authors

Piers Dunstan and Scott Foster

Details

Fits a selected distribution to the vector of relative abundances. See Foster and Dunstan 2009 for details.

Examples

head(n.data)
## format for the data object
n.data$N.scale <- n.data$N/n.data$area
n.data$S.scale <- n.data$S/n.data$area
nij.form <- nij~1 + N.scale + S.scale  + depth + O2_AV + temp_AV
model.nij <- MDMnMod(nij.form, data=n.data, ID=n.data$i, dist="MDMn", scale.covar=TRUE, est.var=FALSE, calc.resid=TRUE, trace=TRUE)
plot(model.nij$fitted$nij,model.nij$residuals)
plot(log(model.nij$fitted$nij),model.nij$residuals)