sdca: Smoothly Detrended Correspondence Analysis

In jarioksa/natto: An Extreme 'vegan' Package of Experimental Code

Smoothly Detrended Correspondence Analysis

Description

Function sdca is similar to decorana, but instead of detrending by segments it uses loess for smooth non-linear detrending.

Usage

sdca(Y, iweigh = FALSE, pairwise = FALSE, monitor = FALSE, ...)

Arguments

Y	Input data.
iweigh	Downweight rare species.
pairwise	Detrend axis k separately for each previous axis in order 1 \dots k-1 \dots 1. This only influences axes 3 and 4.
monitor	Turn on graphical monitoring of detrending for each axis.
...	Other arguments (passed to loess).

Details

Detrended correspondence analysis (DCA) as implemented in decorana tries to remove all systematic biases between ordination axes by detrending later axes against previous ones (Hill & Gauch 1980). DCA uses an ingenious method of detrending by axis segments to allow removing non-linear dependencies. Detrending means taking residuals against smoothed segment means as the new ordination scores for the current axis. It has been suggested that abrupt changes at segment borders can cause some problems in DCA. The current function replaces segmented detrending with detrending against smooth loess functions. However, in many cases this changes the results little from the original detrending by segments.

The detrending for axes 3 and 4 is perfomed either using all previous axes simultaneously in loess (default) or if pairwise=TRUE performing sequential separate detrending against each previous axis separately so that both the first and last detrending are performed against axis 1 similarly as in the original decorana. For axis 3 the sequence is against axes 1, 2, 1 and for axis 4 against axes 1, 2, 3, 2, 1.

The decorana software made several other innovations than detrending. Rescaling of axes is often more influential in application than the actual detrending, like you can see by using decorana without rescaling.

Value

Function returns a subset of decorana result object and can use many decorana methods (such as plot).

References

Hill, M.O. and Gauch, H.G. (1980). Detrended correspondence analysis: an improved ordination technique. Vegetatio 42, 47–58.

Examples

data(spurn)
mod <- sdca(spurn)
plot(mod, display="species")
if (require(vegan)) {
## compare against original decorana without rescaling
mod0 <- decorana(spurn, iresc = 0)
plot(procrustes(mod0, mod, choices=1:2))
}

jarioksa/natto documentation built on March 28, 2024, 12:45 a.m.