indval: (Blocked) Indicator Species Analysis

In phytomosaic/ecole: ecole: School of Ecology Package

Description

Blocked or unblocked versions of Dufrene-Legendre indicator species analysis. If blocked, then can also calculate conditional probabilities of gain/loss/stasis using Bayes rule. If blocked, then blocks are assumed to be sample units and groups are assumed to be the two sample events.

Usage

indval(spe, grp, blk = NULL, cond_probs = TRUE, ...)

Arguments

spe	array, where rows = SUs and cols = species. Values may be numeric abundances or presence/absence.
grp	vector, describing group membership (required).
blk	vector, describing block membership (optional).
cond_probs	logical, should conditional probabilities be calculated when blocked structure exists? Default 'TRUE'.
...	further arguments currently ignored.

Details

Unblocked version follows Dufrene and Legendre (1997). Blocked indicator species analysis was first available in PC-ORD (McCune and Mefford 2017). For blocked designs, the current function can also calculate conditional probabilities of species gains, losses, and stasis across two sampling events. Conditional probabilities use Bayes' rule.

Value

List containing items:
- cond_probs: conditional probabilities, when blocks exist.
- A: relative abundance component of IndVal per group.
- B: relative frequency component of IndVal per group.
- IV: IndVal, the indicator value per group.
- IVmax: maximum IndVal in the indicated group.
- IVmax_dir: signed max IndVal, when blocks exist.
- maxgrp: the group in which each species attains max IndVal.
- sumIVmax: sum of max IndVal across all species.
- sig_expected: number of significant indicator species expected at random.
- pval: permutation p-values for each species.

References

Dufrene, M. and Legendre, P. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs 67(3):345-366.

McCune, B., and M. J. Mefford. 2017. PC-ORD. Multivariate Analysis of Ecological Data. Version 7. MjM Software Design, Gleneden Beach, OR.

See Also

labdsv::indval or indicspecies::strassoc

Examples

data(braun)
spe <- braun$spe[-1,]                 # force an even number

# no group structure exists, so lets pretend we have revisit data
blk <- rep(1:(NROW(spe)/2), times=2)  # 173 sites 'visited' twice
grp <- rep(c(1,2), ea=NROW(spe)/2)    # 'visits' 1 or 2 for each site

# unblocked
iv  <- indval(spe, grp)

# blocked by site
ivb <- indval(spe, grp, blk, cond_probs=TRUE)

# see effects of blocking
plot(iv$IVmax, ivb$IVmax) ; abline(0,1)
table(unblocked = iv$maxgrp, blocked = ivb$maxgrp)

# given our haphazard group assignment, most species have poor indicator value
hist(ivb$IVmax, breaks=22, xlab='Max IndVal', main='')
iv$sig_expected   # num indicator species expected at random

# see how IndVal compares to conditional gain/loss probabilities blocked by site
a <- data.frame(ivb$cond_probs)
plot(a$prloss, a$prgain, col=ecole::colvec(ivb$IVmax_dir), pch=16)

phytomosaic/ecole documentation built on Jan. 2, 2022, 11:24 p.m.