# plot.manyany: Plot Diagnostics for a manyany or glm1path Object In mvabund: Statistical Methods for Analysing Multivariate Abundance Data

## Description

A residual vs fits plot from a `manyany` or `glm1path` object.

## Usage

 ```1 2``` ```## 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` or `glm1path`. 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.

Note also that for `glm1path` objects, slope coefficients have been shrunk towards zero so it is not unusual to see an increasing slope on the residual plot, with larger residuals coinciding with larger fitted values. This arises a a consequent of shrinkage, check if it goes away upon removing the penaly term (e.g. on refitting using `manyglm`) before ringing any alarm bells.

## Author(s)

David Warton <[email protected]>.

## 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.

`glm1path`, `manyany`, `manyglm`

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16``` ```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... ```

mvabund documentation built on May 29, 2017, 9:50 a.m.