swapClass (version 1.0.1)

nullModel: SwapClass null model

Description

a simple null model adapted to semi-quantitative multi-classes (or partially ordered) data.

Usage

nullModel(mat, nsim = 100, recursive = TRUE, burnin = NULL, thin = NULL, strata = NULL)

Arguments

mat

an object of class matrix (or data.frame) with numerical abundance classes (site by row and species by column).

nsim

Number of simulated null communities (positive integer), default = 100.

recursive

A logical value indicating if generation of a new null community is performed from the last one created, defaut = T.

burnin

Number of null communities discarded before proper analysis (positive integer), if NULL (defaut) then burnin = nrow(mat)*ncol(mat)*10.

thin

Number of discarded null communities between two evaluations (positive integer) if recursive=T, if NULL (defaut) then thin = max(nrow(mat)*ncol(mat)*10,1000).

strata

a numeric vector of length egal to nrow(mat) supplying strata. Swap are performed within strata, defaut = NULL.

Value

sim

an object of class list containing the null communities

perms

an object of class list containing matrices with the number of times each cell has been swapped.

Details

The SwapClass model is derived from the "swap philosophy" apply on presence-absence data. The limited number of abundance classes and their repetitiveness allow the extension of the "swap" method classically applied on presence-absence data with two classes (0 and 1). In the same way that "swap" methods permute sub-matrices of presence/absence community matrix, semi-quantitative data can be randomized by swapping sub-matrices while row and column marginals are not modified. nullModel calls a dynamically defined functions (swapClass_swapC) with in-lined C code compiled by setCMethod on package load.

References

Borgy B. - Dynamic and assembly of weed communities: Approach by statistical modeling (2011). Ph.D. Thesis. INRA Agroicologie & Universiti de Bourgogne. DOI: 10.13140/RG.2.1.1738.1601

See Also

nbPerm

Examples

MAT=matrix(sample(0:3,50*20,replace=TRUE,prob=c(.7,.1,.1,.1)),ncol=20)

#Generation of null communities
NULL_MATS=nullModel(MAT)

#First null community
NULL_MATS$sim[[1]]

#Number of times that each cell has been swapped for the first null community
NULL_MATS$perms[[1]]

#nbPerm index over the observed community and the 100 null communities
plot(c(nbPerm(MAT),unlist(lapply(NULL_MATS$sim,nbPerm))),type='l',ylab="nbPerm index")

#the number of each classes per row are equal
#between observed community and the first null community
f_table = function(x) table(factor(x,levels=0:3))
all(apply(MAT,1,f_table)==apply(NULL_MATS$sim[[1]],1,f_table))

#the number of each classes per column are equal
#between observed community and the first null community
all(apply(MAT,2,f_table)==apply(NULL_MATS$sim[[1]],2,f_table))