kdist {ade4}

the class of objects 'kdist' (K distance matrices)

Description

An object of class kdist is a list of distance matrices observed on the same individuals

Usage

kdist(..., epsi = 1e-07, upper = FALSE)

Arguments

...

a sequence of objects of the class kdist.

epsi

a tolerance threshold to test if distances are Euclidean (Gower's theorem) using \frac{\lambda_n}{\lambda_1} is larger than -epsi.

upper

a logical value indicating whether the upper of a distance matrix is used (TRUE) or not (FALSE).

Details

The attributs of a 'kdist' object are:
names: the names of the distances
size: the number of points between distances are known
labels: the labels of points
euclid: a logical vector indicating whether each distance of the list is Euclidean or not.
call: a call order
class: object 'kdist'

Value

returns an object of class 'kdist' containing a list of semidefinite matrices.

Author(s)

Daniel Chessel
Anne-Béatrice Dufour [e-mail]

References

Gower, J. C. (1966) Some distance properties of latent root and vector methods used in multivariate analysis. Biometrika, 53, 325–338.

Examples

# starting from a list of matrices 
data(yanomama)
lapply(yanomama,class)  
kd1 = kdist(yanomama)
print(kd1)

# giving the correlations of Mantel's test
cor(as.data.frame(kd1))
pairs(as.data.frame(kd1))

# starting from a list of objects 'dist'
data(friday87)
fri.w <- ktab.data.frame(friday87$fau, friday87$fau.blo, 
    tabnames = friday87$tab.names)
fri.kd = lapply(1:10, function(x) dist.binary(fri.w[[x]],2))
names(fri.kd) = friday87$tab.names
unlist(lapply(fri.kd,class)) # a list of distances
fri.kd = kdist(fri.kd)
fri.kd
s.corcircle(dudi.pca(as.data.frame(fri.kd), scan = FALSE)$co)

# starting from several distances
data(ecomor)
d1 <- dist.binary(ecomor$habitat, 1)
d2 <- dist.prop(ecomor$forsub, 5)
d3 <- dist.prop(ecomor$diet, 5)
d4 <- dist.quant(ecomor$morpho, 3)
d5 <- dist.taxo(ecomor$taxo)
ecomor.kd <- kdist(d1, d2, d3, d4, d5)
names(ecomor.kd) = c("habitat", "forsub", "diet", "morpho", "taxo")
class(ecomor.kd)
s.corcircle(dudi.pca(as.data.frame(ecomor.kd), scan = FALSE)$co)

data(bsetal97)
X <- prep.fuzzy.var(bsetal97$biol, bsetal97$biol.blo)
w1 <- attr(X, "col.num")
w2 <- levels(w1)
w3 <- lapply(w2, function(x) dist.quant(X[,w1==x], method = 1))
names(w3) <- names(attr(X, "col.blocks"))
w3 <- kdist(list = w3)
s.corcircle(dudi.pca(as.data.frame(w3), scan = FALSE)$co)

data(rpjdl)
w1 = lapply(1:10, function(x) dist.binary(rpjdl$fau, method = x))
w2 = c("JACCARD", "SOKAL_MICHENER", "SOKAL_SNEATH_S4", "ROGERS_TANIMOTO")
w2 = c(w2, "CZEKANOWSKI", "S9_GOWER_LEGENDRE", "OCHIAI", "SOKAL_SNEATH_S13")
w2 <- c(w2, "Phi_PEARSON", "S2_GOWER_LEGENDRE")
names(w1) <- w2
w3 = kdist(list = w1)
w4 <- dudi.pca(as.data.frame(w3), scan = FALSE)$co
w4