witwit.model: Within Correspondence Analysis using divers Models and...

In pamctdp: Principal Axes Methods for Contingency Tables with Partition Structures on Rows and Columns

Description

witwit.model performs an Double Within Tables Correspondence Analysis. Modification of witwit.coa of ade4 to allow Intra Block Model and divers weights

Usage

witwit.model(dudi, row.blocks, col.blocks, pfil = dudi$lw, pcol = dudi$cw, 
    model = "C", weight = "coa", scannf = TRUE, nf = 2,eps=1e-15,iter=100)
## S3 method for class 'wwmodel'
summary(object, ...) 
## S3 method for class 'wwmodel'
print(x, ...)   

Arguments

dudi	an object of class coa
row.blocks	a numeric vector indicating the row numbers for each block of rows
col.blocks	a numeric vector indicating the column numbers for each block of columns
scannf	a logical value indicating whether the eigenvalues bar plot should be displayed
nf	if scannf FALSE, an integer indicating the number of kept axes
pfil	a numeric vector indicating the row weights
pcol	a numeric vector indicating the column weights
model	"C": the same model of ICA,"B": intra blocks independence model
weight	c("coa": the same row and columns weights than CA, "mfa": MFA-like weights in rows and columns, "mafc": MFA-like weights in columns, "mfar": MFA-like weights in rows
eps	convergence error if weight="mfa"
iter	maximum itection number if if weight="mfa"

object

an object of class wwmodel

x	an object of class wwmodel
...	further arguments passed to or from other methods

Details

This function is build up with witwit.coa of ade4, in order to allow diferents weights and models in a contingency table with double structure of partition. If model="C" and weight="coa" the results are the same of witwit.coa. If model="B" and weight="coa" a Intra-Blocks Correspondence Analysis (IBCA) is buld up If model="B" and weight="mfa" a Weighted Intra-Blocks Correspondence Analysis (WIBCA) is build up

Value

Returns a list of class wwmodel containing:

tab	a data frame with I rows and K columns
cw	column weights, a vector with K components
lw	row weights, a vector with I components
eig	eigenvalues, a vector with min(I,K) components
nf	integer, number of kept axes
c1	principal axes, data frame with I rows and nf columns
l1	principal components, data frame with I rows and nf columns
co	column coordinates, data frame with K rows and nf columns
li	row coordinates, data frame with I rows and nf columns
call	original call
rbvar	a data frame with the within variances of the rows of the factorial coordinates
lbw	a data frame with the marginal weighting of the row bands
cvar	a data frame with the within variances of the columns of the factorial coordinates
cbw	a data frame with the marginal weighting of the column bands
hom	homotecia to read some aids as in MFA
rbl	number of rows in each row-band
cbl	number of columns in each column-band
sepeig.col	band-column separate firt eigenvalues if weight="mfa"
sepeig.row	band-row separate firt eigenvalues if weight="mfa"

Author(s)

Campo Elías PARDO [e-mail]

References

Becue M., Pages J. and Pardo C.E. (2005). Contingency table with a double partition on rows and columns. Visualization and comparison of the partial and global structures. In: Proceedings ASMDA, Brest, France. May,17-20, 2005. Eds: Jacques Janssen and Philippe Lenca. ENST Bretagne. pages 355–364. http://conferences.telecom-bretagne.eu/asmda2005/IMG/pdf/proceedings/355.pdf

Cazes, P., Chessel, D. and Doledec, S. (1988) L'analyse des correspondances internes d'un tableau partitionne : son usage en hydrobiologie. Revue de Statistique Appliquee, 36, 39–54. http://archive.numdam.org/ARCHIVE/RSA/RSA_1988__36_1/RSA_1988__36_1_39_0/RSA_1988__36_1_39_0.pdf

Pardo, Campo Elías, Mónica Bécue-Bertaut, and Jorge Eduardo Ortiz. (2013). Correspondence Analysis of Contingency Tables with Subpartitions on Rows and Columns. Revista Colombiana de Estadística 36.1: 115–144.

Examples

data(ardeche)
# change column names
names(ardeche$tab) <- paste(ardeche$sta.fac,ardeche$dat.fac,sep="") 
rownames(ardeche$tab) <- # change row names
paste(strtrim(rownames(ardeche$tab),1),substr(rownames(ardeche$tab),4,
	length(rownames(ardeche$tab))),sep="")
coa1 <- dudi.coa(ardeche$tab, scannf = FALSE, nf = 4)
ww <- witwit.model(coa1, ardeche$row.blocks, ardeche$col.blocks, scann = FALSE)
ww
plot(ww)
summary(ww)

pamctdp documentation built on May 1, 2019, 10:19 p.m.