residLen {analogue}

Squared residual length diagnostics

Description

The squared residual length between the fitted values of a constrained ordination and the original species data is one diagnostic for transfer function models.

Usage

residLen(X, env, passive, method = c("cca","rda"))

fittedY(ord, newdata, colsum)

sqrlLinear(Y, fitted)

sqrlUnimodal(Y, colsum, fitted)

Arguments

X

data frame; the training set species data.

env

vector; the training set environmental data.

passive

data frame; the passive samples species data.

method

the ordination technique to use. One of "rda" or "cca", with the latter the default.

ord

an ordination object, the result of a call to cca or rda.

newdata

Species data matrix for passive samples. Must have same columns as data used to fit ord.

colsum

column (species) sums for training set data used to fit ord.

Y

Original species data matrix, the response for which squared residual lengths are to be computed.

fitted

The fitted values of the response derived from the constrained ordination model.

Details

The squared residual lengths are computed for the training set samples and the passive samples separately. Passive samples that are poorly fitted in the transfer function model will have large squared residual distances between the observed species data and the fitted values from the constrained ordination.

residLen is the main user-interface function and can be called with either the training data and passive samples.

fittedY returns the fitted approximation of the passive sample response data (i.e. species data). sqrlLinear and sqrlUnimodal return the squared residual distances between the observed species data and the fitted values from the constrained ordination model.

Value

fittedY returns a matrix of fitted species abundances for passive samples.

sqrlLinear and sqrlUnimodal return a vector of residual distances.

residLen returns an object of class "residLen" with the attribute "method" set to "method". This object is a list with the following components:

train, passive

The squared residual lengths for the training set and the passive samples.

ordination

The fitted ordination.

call

The matched call.

Author(s)

Gavin L. Simpson

References

Ter Braak C.J.F. and Smilauer P. (2002) CANOCO Reference manual and CanoDraw for Windows User's guide: Software for Canonical Ordination (version 4.5). Microcomputer Power (Ithaca, NY, USA), 500pp.

See Also

cca and predict.cca for some of the underlying computations.

Examples

## load the Imbrie and Kipp example data
data(ImbrieKipp, SumSST, V12.122)

## squared residual lengths for Core V12.122
rlens <- residLen(ImbrieKipp, SumSST, V12.122)
rlens

## as before but using linear RDA
residLen(ImbrieKipp, SumSST, V12.122, method = "rda")