list2CubeOfCovDis function

compute a cube of covariance and a cube of distance between the items (rows) of a matrix of measurements comprising $K$ different blocks of possibly different number of variables.

list2CubeOfCovDis

compute a cube of covariance and a cube of (squared) Euclidean distance between the items (rows) a matrix of measurements comprising $K$ different blocks of possibly different number of variables. The variables describing the items can scaled to norm 1 and centered. The whole matrix for a block can be scaled by its first eigenvalue (a la DISTATIS). Blocks can have different number of variables and when all blocks have same number of variables list2CubeOfCovDis

is a more efficient alternative

list2CubeOfCovDis(Data, Judges, scale = TRUE, center = TRUE, ev.scale = TRUE)

Arguments

• Data: a matrix of dimensions $I$ items by $J$ quantitative variables (structured in $K$ blocks of $J_k$ variables each). No Default.

• Judges: a $J$ components character vector identifiying the variables corresponding to each block of variables. No Default.

• scale: (Default: TRUE), when TRUE

  scale to norm 1 each column for each slice.

• center: (Default: TRUE), when TRUE

  centers each column.

• ev.scale: (Default: TRUE), when TRUE

  normalizes each slice (i.e., each $I$ items by $J$ matrix) so that its first eigenvalue is equal to 1.

Returns

a list with 1) cubeOfCovariance

a cube of $K$ $I$ by $I$ covariance matrices; and 2) codecubeOfDistance

a cube of $K$ $I$ by $I$

(squared) Euclidean distance matrices.

Details

The input of list2CubeOfCovDis is a $I$ items by $J$ quantitative variables that are organized in $K$ blocks (i.e., submatrices) each comprising $J_k$ variables (with sum $J_k = J$).

By default list2CubeOfCovDis

centers and normalizes each column for each block and then normalize each covariance matrix such that the first eigenvalue of each covariance matrix (for a given block) is equal to 1.

A distatis analysis of the Distance matrices with the option Distance = TRUE will give the same results as the distatis analysis of the Covariance matrices with the option Distance = FALSE.

Examples

path2file <- system.file("extdata",
                        "BeersFlashProfile.xlsx", 
                        package = 'DistatisR')
# read the data in the excel file with read.df.excel
beerDataFlash  <- read.df.excel(path = path2file,
                               sheet = 'Rankings')$df.data
# Extract the namers of the judges (first 2 characters)
JudgesVars <- colnames(beerDataFlash)
zeJudges <- substr(JudgesVars,1,2)
 # call list2CubeOfCovDis
test.list2 <- list2CubeOfCovDis(Data = beerDataFlash ,
                               Judges =  zeJudges)

See Also

list2CubeOfCov

Author(s)

Herve Abdi