adiv (version 2.2.1)

EqRao: Apportionment of Diversity

Description

Eight functions are available.

EqRS performs the first decomposition of diversity developed in Pavoine et al. (2016) and rtestEqRS can be used for the associated permutation tests.

EqRSintra performs the second decomposition of diversity developed in Pavoine et al. (2016) and rtestEqRSintra can be used for the associated permutation tests.

EqRao performs the third decomposition of diversity introduced in Pavoine et al. (2016) and rtestEqRao can be used for the associated permutation tests.

wapqe performs the apportionment of quadratic entropy (Rao 1986) and rtestwapqe associated permutation tests (Pavoine et al. 2016).

Usage

EqRS(comm, dis = NULL, structures = NULL, 
option = c("eq", "normed1", "normed2"), 
formula = c("QE", "EDI"), tol = 1e-08)
EqRSintra(comm, dis = NULL, structures = NULL, 
option = c("eq", "normed1", "normed2"), 
formula = c("QE", "EDI"), tol = 1e-08, 
metmean = c("harmonic", "arithmetic"))
EqRao(comm, dis = NULL, structures = NULL, 
option = c("eq", "normed1", "normed2"), 
formula = c("QE", "EDI"), wopt = c("even", "speciesab"), 
tol = 1e-08, metmean = c("harmonic", "arithmetic"))
wapqe(comm, dis = NULL, structures = NULL, 
formula = c("QE", "EDI"), wopt = c("even", "speciesab"), 
tol = 1e-08)
rtestEqRS(comm, dis = NULL, structures = NULL, 
formula = c("QE", "EDI"), 
option = c("normed1", "normed2", "eq"), 
popt = c("aggregated", "independent"), level = 1, nrep = 99, 
alter = c("greater", "less", "two-sided"), tol = 1e-08)
rtestEqRSintra(comm, dis = NULL, structures = NULL, 
formula = c("QE", "EDI"), 
option = c("normed1", "normed2", "eq"), 
popt = c("aggregated", "independent"), level = 1, 
nrep = 99, alter = c("greater", "less", "two-sided"), 
tol = 1e-08, metmean = c("harmonic", "arithmetic"))
rtestEqRao(comm, dis = NULL, structures = NULL, 
formula = c("QE", "EDI"), 
option = c("normed1", "normed2", "eq"), 
wopt = c("even", "speciesab"), 
popt = c("aggregated", "independent"), level = 1, nrep = 99, 
alter = c("greater", "less", "two-sided"), tol = 1e-08, 
metmean = c("harmonic", "arithmetic"))
rtestwapqe(comm, dis = NULL, structures = NULL, 
formula = c("QE", "EDI"), wopt = c("even", "speciesab"), 
popt = c("aggregated", "independent"), 
level = 1, nrep = 99, alter = c("greater", "less", "two-sided"),
tol = 1e-08)

Value

The functions EqRS, EqRSintra, EqRao, wapqe return a data frame with each component of the selected diversity decomposition. The functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe return a list of class randtest or krandtest (classes of package ade4).

Arguments

comm

a data frame or a matrix with communities as rows and species as columns. Entries are abundances of species within sites.

dis

either NULL or an object of class dist that contains the (functional or phylogenetic) distances among species. If NULL, the Gini-Simpson index is used.

structures

either NULL or a data frame that contains, in the jth row and the kth column, the name of the group of level k to which the jth community belongs. Communities in structures should be in the same order as in comm. See details.

option

"eq", "normed1" or "normed2" (if a vector with several codes is given, only the first one is considered). "eq" means that the diversity components are given in terms of equivalent number (E) of species, sites, regions etc. "normed1" means that the normed components of diversity will be returned with formula (1 - 1 / E) / (1 - 1 / Emax); "normed2" means that the normed components of diversity will be returned with formula (E - 1) / (Emax - 1). Emax is the maximum possible value for E. For Ealpha (local diversity) and Egamma (global diversity), Emax=S (the number of species in the data set).

formula

either "QE" (default) or "EDI". See details.

wopt

either "even" (default), "speciesab", or a numeric vector. If w="speciesab", then the communities will be weighted by their sum of species' abundances. If w="even", the communities will be evenly weighted within the factors defined by the argument structures.

popt

a sring: either "aggregated" or "independent". This choice will define the permutation test within communities (level 1 of the permutation scheme). If "aggregated" is chosen, sampled individuals will be considered to be aggregated in space so that. (For example, consider that fish are sampled by net, and fish are aggregated in space, then when using the net, one might have zero abundance once because the net was not placed in a shoal of fish and a very high abundance another time because the net was placed in a shoal of fish; the permutation scheme was developed for this scenario). In that case, the abundance values of each species are permuted between communities. Permutations leading to an empty community (with all species abundances equal to zero) are excluded. If "independent" is chosen, sampled individuals will be considered independent from each other and now abundances are randomly defined with the only constraints that the sum of species abundance within a community is kept constant and the sum of abundance of a species across all communities (if argument structures is NULL or across the communities of a group (of level 2, see argument level) if structures is not NULL. Decimals are not allowed for abundance data with popt = "independent".

tol

a tolerance threshold (a value between -tol and tol is considered equal to zero).

metmean

a string: either "arithmetic" or "harmonic" (default).

level

a number. The number is discarded if the argument structures is set to NULL. If structures is different from NULL then 1 means test for differences among communities, within the levels of the first factor given in argument structures (column 1), 2 means test for differences among levels of the first factor given in argument structures (column 1) but within levels of the second factor given in argument structures (column 2) (if available), etc. level should thus be between 1 and 1+the number of columns in structures. See details for concrete examples.

nrep

the number of permutations.

alter

a string specifying the alternative hypothesis, must be one of "greater" (default), "less" or "two-sided".

Author

Sandrine Pavoine

Details

If formula = "QE", the definition of the quadratic entropy is: $$QE(\mathbf{p}_i,\mathbf{D})=\sum_{k=1}^S\sum_{l=1}^S p_{k|i}p_{k|j}d_{kl}$$ where \(\mathbf{p}_i=(p_{1|i}, ..., p_{k|i}, ..., p_{S|i})\) is the vector of relative species abundance within community i; S is the number of species; \(\mathbf{D}=(d_{kl})\) is the matrix of (phylogenetic or functional) dissimilarities among species, and \(d_{kl}\) is the (phylogenetic or functional) dissimilarity between species k and l.

If formula = "EDI", the definition of the quadratic entropy is: $$EDI(\mathbf{p}_i,\mathbf{D})=\sum_{k=1}^S\sum_{l=1}^S p_{k|i}p_{k|j}\frac{d_{kl}^2}{2}$$ EDI stands for the Euclidean Diversity Index of Champely and Chessel (2002). For example, applying EqRao with the distances dis=mydis and formula=QE corresponds to applying it with dis=sqrt(2*mydis) and formula=EDI.

The argument level permits to choose the component of beta diversity that will be tested for significance.

Examples are given below:

Scenario 1: Imagine that structures is NULL, then we only have a set of communities, species within them and a matrix of (functional or phylogenetic) dissimilarities among species. Then there is only one level of beta diversity which represents the average dissimilarities among communities. The functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will always, in that case, test for the significance of the dissimilarities among communities (whatever the value given to argument level).

Scenario 2: Imagine that structures is a data frame with sites as rows and only one column representing how sites are distributed among regions. Then, there are two levels of beta diversity: beta1 diversity represents the (functional or phylogenetic) dissimilarities among sites within regions; and beta2 diversity represents the (functional or phylogenetic) dissimilarities among regions. If level = 1 then functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among sites within regions (beta1 diversity); in contrast, if level = 2 functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among regions (beta2 diversity). As there is only one column in argument structures and thus only two levels of diversity, level cannot be higher than 2.

Scenario 3: Imagine now that argument comm contains quadrats as rows and species as columns and that structures is a data frame with quadrats as rows and three columns representing how quadrats are organized in three nested factors: hamlets, towns, and counties. Then there would be four levels of beta diversity: beta1 diversity represents the (functional or phylogenetic) dissimilarities among quadrats within hamlets, towns, and counties; beta2 diversity represents the dissimilarities among hamlets within towns and counties; beta3 diversity represents the dissimilarities among towns within counties; and beta4 diversity represents the dissimilarities among counties. If level = 1 then functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among quadrats within hamlets, towns, and counties (beta1 diversity); if level = 2 functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among hamlets within towns and regions (beta2 diversity); if level = 3 functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among towns within regions (beta3 diversity); if level = 4 functions rtestEqRS, rtestEqRSintra, rtestEqRao, and rtestwapqe will test for the significance of the dissimilarities among regions (beta4 diversity). As there are only three columns in argument structures and thus only four levels of diversity, level cannot be higher than 4. Test for level=1 is performed by permuting the abundances of each species across quadrats but within hamlets, towns, and counties. Test for level=2 is performed by permuting the quadrats among hamlets within towns and counties. Test for level=3 is performed by permuting the hamlets among towns within counties. Test for level=4 is performed by permuting the towns among counties.

Other permutation schemes will be added in the future.

References

Pavoine, S., Marcon, E., Ricotta, C. (2016) "Equivalent numbers" for species, phylogenetic, or functional diversity in a nested hierarchy of multiple scales. Methods in Ecology and Evolution, 7, 1152--1163.

Champely, S. and Chessel, D. (2002) Measuring biological diversity using Euclideanmetrics. Environmental and Ecological Statistics, 9, 167--177.

Rao, C.R. (1986) Rao's axiomatization of diversity measures. In: Encyclopedia of Statistical Sciences. Vol. 7 (eds S. Kotz and N.L. Johnson), pp. 614--617. New York: Wiley and Sons.

Examples

if (FALSE) {
if(require(ade4)){

data(macroloire, package="ade4")

# Taxonomic dissimilarities among species:
dTaxo <- dist.taxo(macroloire$taxo)^2/2
dTaxo <- dTaxo/max(dTaxo)
# Size-based dissimilarities among species
dSize <- dist.prop(macroloire$traits[ ,1:4], method = 2)
# Dissimilarities among species in terms of feeding categories
dFeed <- dist.prop(macroloire$traits[ ,5:11], method = 2)
# Dissimilarities among species in terms of 
# both size and feeding categories
dSF <- (dSize+dFeed)/2

# Table with sites as rows (stations), 
# species as columns and abundances as entries
ab <- as.data.frame(t(macroloire$fau))
# Table with sites as rows and one column only. 
# Entries indicate the geological region associated 
# with each site
stru <- macroloire$envir["Morphoregion"]

EqRao(ab, , stru, option="eq")
EqRao(ab, dTaxo, stru, formula = "QE", option="eq")
EqRao(ab, dSize, stru, formula = "QE", option="eq")
EqRao(ab, dFeed, stru, formula = "QE", option="eq")
EqRao(ab, dSF, stru, formula = "QE", option="eq")

EqRao(ab, , stru, option="normed2")
EqRao(ab, dTaxo, stru, formula = "QE", option="normed2")
EqRao(ab, dSize, stru, formula = "QE", option="normed2")
EqRao(ab, dFeed, stru, formula = "QE", option="normed2")
EqRao(ab, dSF, stru, formula = "QE", option="normed2")

# Tests for dissimilarities among sites within regions:
### TIME CONSUMING

rb1_GS <- rtestEqRao(ab, , stru, level=1, nrep=999, option="normed2")
rb1_GS
plot(rb1_GS)
rb1_Taxo <- rtestEqRao(ab, dTaxo, stru, formula = "QE", 
level=1, nrep=999, option="normed2")
rb1_Taxo
plot(rb1_Taxo)
rb1_Size <- rtestEqRao(ab, dSize, stru, formula = "QE", 
level=1, nrep=999, option="normed2")
rb1_Size
plot(rb1_Size)
rb1_Feed <- rtestEqRao(ab, dFeed, stru, formula = "QE", 
level=1, nrep=999, option="normed2")
rb1_Feed
plot(rb1_Feed)
rb1_SF <- rtestEqRao(ab, dSF, stru, formula = "QE", 
level=1, nrep=999, option="normed2")
rb1_SF
plot(rb1_SF)

# Tests for dissimilarities among regions:
### TIME CONSUMING

r2_GS <- rtestEqRao(ab, , stru, level=2, nrep=999, option="normed2")
r2_GS
plot(r2_GS)
r2_Taxo <- rtestEqRao(ab, dTaxo, stru, formula = "QE", 
level=2, nrep=999, option="normed2")
r2_Taxo
plot(r2_Taxo)
r2_Size <- rtestEqRao(ab, dSize, stru, formula = "QE", 
level=2, nrep=999, option="normed2", w="even")
r2_Size
plot(r2_Size)
r2_Feed <- rtestEqRao(ab, dFeed, stru, formula = "QE", 
level=2, nrep=999, option="normed2", w="even")
r2_Feed
plot(r2_Feed)
r2_SF <- rtestEqRao(ab, dSF, stru, formula = "QE", 
level=2, nrep=999, option="normed2", w="even")
r2_SF
plot(r2_SF)
}
}