utils: Postprocessing: renormalization, reflection and reordering;...

In rrcov3way: Robust Methods for Multiway Data Analysis, Applicable also for Compositional Data

Postprocessing: renormalization, reflection and reordering; access to some of the components of the model.

Description

The estimated model will be renormalized, reflected (change of sign) or the components will be reordered. Functions that provide access to some components of the model: coordinates, weights.

Usage

    ## S3 method for class 'tucker3'
do3Postprocess(x, reflectA, reflectB, reflectC, reorderA, reorderB, reorderC, ...)
    ## S3 method for class 'parafac'
do3Postprocess(x, reflectA, reflectB, reflectC, reorder, ...)
    ## S3 method for class 'parafac'
coordinates(x, mode = c("A", "B", "C"), type = c("normalized", "unit", "principal"), ...)
    ## S3 method for class 'tucker3'
coordinates(x, mode = c("A", "B", "C"), type = c("normalized", "unit", "principal"), ...)
    ## S3 method for class 'parafac'
weights(object, ...)
    ## S3 method for class 'tucker3'
weights(object, mode = c("A", "B", "C"), ...)
    ## S3 method for class 'parafac'
reflect(x, mode = c("A", "B", "C"), rsign = 1, ...)
    ## S3 method for class 'tucker3'
reflect(x, mode = c("A", "B", "C"), rsign = 1, ...)

Arguments

x	Tucker3 or Parafac solution
object	Tucker3 or Parafac solution (alternative of x for the generic function weights())
reflectA	How to handle the signs of the components of mode A - can be a single number or a vector with length of the number of components of A
reflectB	How to handle the signs of the components of mode B - can be a single number or a vector with length of the number of components of B
reflectC	How to handle the signs of the components of mode C - can be a single number or a vector with length of the number of components of C
reorder	How to reorder the components of a Parafac solution - a vector with length of the number of components
reorderA	How to reorder the components of mode A - a vector with length of the number of components of A giving the new order
reorderB	How to reorder the components of mode B - a vector with length of the number of components of B giving the new order
reorderC	How to reorder the components of mode C - a vector with length of the number of components of C giving the new order
mode	For which mode to provide the coordinates or weights. Default is mode A
type	Which type of coordinates to provide. Default is "normalized"
rsign	How to change the sign of the components of the given mode. Can be a single number or a vector with length of the number of components of the corresponding mode.
...	Potential further arguments passed to lower level functions.

Value

The output value of do3Postproces() is the postprocessed solution, Parafac or Tucker3. The output of weights() and coordinates() are the respective values.

Author(s)

Valentin Todorov [e-mail] and Maria Anna Di Palma [e-mail] and Michele Gallo [e-mail]

References

Kroonenberg (2008). Applied multiway data analysis. Wiley series in probability and statistics. Hoboken NJ, Wiley.

Examples

    data(elind)
    x1 <- do3Scale(elind, center=TRUE, scale=TRUE)
    (cp <- Parafac(x1, ncomp=3, const=c("orth", "none", "none")))
    
    cp$B
    cp1 <- do3Postprocess(cp, reflectB=-1)      # change the sign of all components of B
    cp1$B
    weights(cp1)
    coordinates(cp1)
    coordinates(cp1, type="principal")

    ## Same as above - the centering and scaling is done inside the Parafac procedure
    (cp1 <- Parafac(elind, ncomp=3, const=c("orth", "none", "none"), 
        center=TRUE, scale=TRUE))

    ## Robust estimation with robust scaling with median and mad
    (cp1 <- Parafac(elind, ncomp=3, const=c("orth", "none", "none"), 
        center=median, scale=mad, robust=TRUE))

    ## Robust estimation with robust scaling with median and Qn (Rousseeuw and Croux, 1993)
    require(robustbase)
    (cp1 <- Parafac(elind, ncomp=3, const=c("orth", "none", "none"), 
        center=median, scale=Qn, robust=TRUE))

rrcov3way documentation built on July 9, 2023, 7:44 p.m.