mvabund (version 3.9.1)

plot.manyany: Plot Diagnostics for a manyany Object

Description

A residual vs fits plot from a manyany object.

Usage

## S3 method for class 'manyany':
plot( x, ...)

Arguments

x

manyany object, resulting from a call to manyany.

...

other parameters to be passed through to plotting functions.

Details

plot.manyany is used to check assumptions that are made when fitting a model via manyany. As in Wang et al (2012), you should check the residual vs fits plot for no pattern (hence no suggestion of failure of any linearity and mean-variance assumptions). It is also desirable that residuals follow a straight line of slope one on a normal Q-Q plot. These plots use Dunn-Smyth residuals (Dunn & Smyth 1996), described at residuals.manyglm. Note that for discrete data, these residuals involve random number generation, and will not return identical results on replicate runs - so it is recommended that you plot your data a few times to check if any pattern shows up consistently across replicate plots.

References

Dunn, P.K., & Smyth, G.K. (1996). Randomized quantile residuals. Journal of Computational and Graphical Statistics 5, 236-244. Wang Y., Naumann U., Wright S.T. & Warton D.I. (2012). mvabund - an R package for model-based analysis of multivariate abundance data. Methods in Ecology and Evolution 3, 471-474.

See Also

manyany

Examples

require(graphics)

data(spider)
abund <- mvabund(spider$abund)
X <- spider$x

## Plot the diagnostics for a log-linear model assuming counts are poisson:
spidPois <- manyany("glm", abund, data=X, abund ~ X, family=poisson())
plot(spidPois,pch=19,cex=0.2)
## Fan-shape means trouble for our Poisson assumption.

## Try a negative binomial instead...
require(MASS) # this package is needed for its negative binomial family function
spidNB <- manyany("manyglm", abund, data=X, abund ~ X, family="negative.binomial")
plot(spidNB,pch=19,cex=0.2,xlim=c(-15,6))
## That's looking a lot better...