permustats: Extract, Analyse and Display Permutation Results

permustatsR Documentation

Extract, Analyse and Display Permutation Results

Description

The permustats function extracts permutation results of vegan functions. Its support functions can find quantiles and standardized effect sizes, plot densities and Q-Q plots.

Usage

permustats(x, ...)
## S3 method for class 'permustats'
summary(object, interval = 0.95, alternative, ...)
## S3 method for class 'permustats'
densityplot(x, data, xlab = "Permutations", ...)
## S3 method for class 'permustats'
density(x, observed = TRUE, ...)
## S3 method for class 'permustats'
qqnorm(y, observed = TRUE, ...)
## S3 method for class 'permustats'
qqmath(x, data, observed = TRUE, sd.scale = FALSE,
    ylab = "Permutations", ...)
## S3 method for class 'permustats'
boxplot(x, scale = FALSE, names, ...)
## S3 method for class 'permustats'
pairs(x, ...)

Arguments

object, x, y

The object to be handled.

interval

numeric; the coverage interval reported.

alternative

A character string specifying the limits used for the interval and the direction of the test when evaluating the p-values. Must be one of "two.sided" (both upper and lower limit), "greater" (upper limit), "less" (lower limit). Usually alternative is given in the result object, but it can be specified with this argument.

xlab, ylab

Arguments of densityplot and qqmath functions.

observed

Add observed statistic among permutations.

sd.scale

Scale permutations to unit standard deviation and observed statistic to standardized effect size.

data

Ignored.

scale

Use standardized effect size (SES).

names

Names of boxes (default: names of statistics).

...

Other arguments passed to the function. In density these are passed to density.default, and in boxplot to boxplot.default.

Details

The permustats function extracts permutation results and observed statistics from several vegan functions that perform permutations or simulations.

The summary method of permustats estimates the standardized effect sizes (SES) as the difference of observed statistic and mean of permutations divided by the standard deviation of permutations (also known as z-values). It also prints the the mean, median, and limits which contain interval percent of permuted values. With the default (interval = 0.95), for two-sided test these are (2.5%, 97.5%) and for one-sided tests either 5% or 95% quantile and the p-value depending on the test direction. The mean, quantiles and z values are evaluated from permuted values without observed statistic, but the p-value is evaluated with the observed statistic. The intervals and the p-value are evaluated with the same test direction as in the original test, but this can be changed with argument alternative. Several permustats objects can be combined with c function. The c function checks that statistics are equal, but performs no other sanity tests.

The density and densityplot methods display the kernel density estimates of permuted values. When observed value of the statistic is included in the permuted values, the densityplot method marks the observed statistic as a vertical line. However the density method uses its standard plot method and cannot mark the observed value.

The qqnorm and qqmath display Q-Q plots of permutations, optionally together with the observed value (default) which is shown as horizontal line in plots. qqnorm plots permutation values against standard Normal variate. qqmath defaults to the standard Normal as well, but can accept other alternatives (see standard qqmath). The qqmath function can also plot observed statistic as standardized effect size (SES) with standandized permutations (argument sd.scale). The permutations are standardized without the observed statistic, similarly as in summary.

Functions density and qqnorm are based on standard R methods and accept their arguments. They only handle one statistic, and cannot be used when several test statistic were evaluated. The densityplot and qqmath are lattice graphics, and can be used either for one or for several statistics. All these functions pass arguments to their underlying functions; see their documentation. Functions qqmath and densityplot default to use same axis scaling in all subplots of the lattice. You can use argument scales to set independent scaling for subplots when this is appropriate (see xyplot for an exhaustive list of arguments).

Function boxplot draws the box-and-whiskers plots of effect size, or the difference of permutations and observed statistic. If scale = TRUE, permutations are standardized to unit standard deviation, and the plot will show the standardized effect sizes.

Function pairs plots permutation values of statistics against each other. The function passes extra arguments to pairs.

The permustats can extract permutation statistics from the results of adonis2, anosim, anova.cca, mantel, mantel.partial, mrpp, oecosimu, ordiareatest, permutest.cca, protest, and permutest.betadisper.

Value

The permustats function returns an object of class "permustats". This is a list of items "statistic" for observed statistics, permutations which contains permuted values, and alternative which contains text defining the character of the test ("two.sided", "less" or "greater"). The qqnorm and density methods return their standard result objects.

Author(s)

Jari Oksanen with contributions from Gavin L. Simpson (permustats.permutest.betadisper method and related modifications to summary.permustats and the print method) and Eduard Szöcs (permustats.anova.cca).

See Also

density, densityplot, qqnorm, qqmath.

Examples

data(dune, dune.env)
mod <- adonis2(dune ~ Management + A1, data = dune.env)
## use permustats
perm <- permustats(mod)
summary(perm)
densityplot(perm)
qqmath(perm)
boxplot(perm, scale=TRUE, lty=1, pch=16, cex=0.6, col="hotpink", ylab="SES")
abline(h=0, col="skyblue")
## example of multiple types of statistic
mod <- with(dune.env, betadisper(vegdist(dune), Management))
pmod <- permutest(mod, nperm = 99, pairwise = TRUE)
perm <- permustats(pmod)
summary(perm, interval = 0.90)

vegan documentation built on Oct. 11, 2022, 5:06 p.m.