resamp_nmds: Resampled NMDS

In phytomosaic/fitNMDS: Fitting Resampled and Reciprocal NMDS Models

Description

Bootstrap or jackknife resampling of nonmetric multidimensional scaling (NMDS) configurations, to estimate internal sampling variability. Accompanied by summary and plot methods.

Usage

resamp_nmds(spe, method = "bray", zeroadj = TRUE, step = TRUE,
  k = 2, BS = NULL, B = 9, ...)

## S3 method for class 'resamp_nmds'
summary(object, ...)

## S3 method for class 'resamp_nmds'
plot(x, noaxes = F, col = "#00000040",
  type = "spider", ...)

Arguments

spe	species abundance matrix
method	dissimilarity index, per vegdist
zeroadj	default TRUE: adjust dissimilarities to account for zero-sum rows?
step	default TRUE: use shortest-path stepacross dissimilarity adjustment?
k	number of dimensions sought in final NMDS solution
BS	bootstrap size (currently constrained to not exceed the number of rows in original matrix). Default BS=NULL, or else BS=1, will perform jackknife NMDS (leave-one-out).
B	number of bootstrap replicates
...	additional arguments passed to function
x, object	result from resamp_nmds
noaxes	default TRUE: cleanly plot without axes?
col	character vector of colors for spider lines
type	plot type, one of 'spider' or 'points' for ordination plots, or 'hist' to plot the distribution of fit statistics across resampling replicates

Details

Bootstrap or jackknife resampling of NMDS with resamp_nmds estimates internal sampling variability of a candidate dataset. The premise of this is resampling with replacement from the dissimilarities matrix, performing NMDS ordination (Kruskal 1964) on each resampled replicate, then determining collective sampling variability across all NMDS solutions as rP_internal. Other uses of resampled NMDS may include estimating confidence regions for site scores, or testing the ‘stability’ of species positions along ordination axes using Scaled Rank Variance of Knox and Peet (1989).

Users may select dissimilarity index per vegdist, as well as options for zero-adjusted dissimilarities (Clarke et al. 2006) using zeroadj=TRUE, and/or shortest-path stepacross dissimilarities using step=TRUE.

Value

List of class ‘resamp_nmds’ with elements:

rP_draw:

Vector of rP_internal values for each replicate.

stresses:

Vector of NMDS stress for each replicate.

SRV:

Scaled rank variance across all replicates.

rP_internal:

Internal sampling variability across all replicates.

points:

Scores of each replicate configuration.

References

Clarke, K. R., P. J. Somerfield, and M. G. Chapman. 2006. On resemblance measures for ecological studies, including taxonomic dissimilarities and a zero-adjusted Bray–Curtis coefficient for denuded assemblages. J Exp Marine Biol and Ecol 330:55–80.
Knox, R. G., and R. K. Peet. 1989. Bootstrapped ordination: a method for estimating sampling effects in indirect gradient analysis. Vegetatio 80:153–165.
Kruskal, J. B. 1964. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika 29: 1-27.

See Also

recip_nmds for reciprocal NMDS, and metaMDS for the core NMDS algorithm

Examples

set.seed(231)
data(smoky)
x <- smoky$spe

# try jackknife and bootstrap on a simplistic dataset
#    (vegan::metaMDS gives warning of near-zero stress)
j <- resamp_nmds(x, k=2)
b <- resamp_nmds(x, B=29, BS=NROW(x), k=2)
summary(j)
summary(b)
par(mfrow=c(1,2))
plot(j)
plot(b)

phytomosaic/fitNMDS documentation built on May 17, 2019, 8:19 p.m.