corReorder {lessR}

Reorder Variables in a Correlation Matrix

Description

Abbreviation: reord

Re-arranges the order of the variables in the input correlation matrix. If no variable list is specified then by default the variables are re-ordered according to hierarchical clustering. Or, re-order with the Hunter (1973) chain method in which the first variable is the variable with the largest sum of squared correlations of all the variables, then the next variable is that with the highest correlation with the first variable, and so forth. Or, re-order manually.

Usage

corReorder(R=mycor, order=c("hclust", "chain", "manual", "as_is"),
          hclust_type = c("complete", "ward.D", "ward.D2", "single",
                          "average", "mcquitty", "median", "centroid"),
          dist_type=c("R", "dist"),
          n_clusters=NULL, vars=NULL, chain_first=0,
          heat_map=TRUE, dendrogram=TRUE, diagonal_new=TRUE,
          main=NULL, bottom=NULL, right=NULL,
          pdf=FALSE, width=5, height=5, ...)

reord(...)

Arguments

R

Correlation matrix.

order

Source of ordering (seriation): Default of hierarchical cluster analysis, Hunter(1973) chain method, manually specified with vars, or left "as_is".

hclust_type

Type of hierarchical cluster analysis.

dist_type

Default is a correlation matrix of similarities, otherwise a distance matrix.

n_clusters

For a hierarchical cluster analysis, optionally specify the cluster membership for the specified number of clusters.

vars

List of the re-ordered variables, each variable listed by its ordinal position in the input correlation matrix. If this is set, then order set to "manual".

chain_first

The first variable listed in the ordered matrix with the chain method.

main

Graph title. Set to main="" to turn off.

heat_map

If TRUE, display a heat map of the item correlations.

dendrogram

If TRUE, display a heat map of the item correlations for a hierarchical cluster analysis.

diagonal_new

If TRUE, replace diagonal for the heat map only with an average of the correlation of item on the diagonal with the two adjacent items.

bottom

Number of lines of bottom margin.

right

Number of lines of right margin.

pdf

Set to TRUE if graphics files written to pdf, the heat map of the re-ordered matrix, and, if an hierarchical cluster analysis, the dendrogram.

width

Width of the pdf file in inches.

height

Height of the pdf file in inches.

...

Parameter values_

Details

Reorder and/or delete variables in the input correlation matrix.

Define the constituent variables, the items, with a listing of each variable by its name in the correlation matrix. If the specified variables are in consecutive order in the input correlation matrix, the list can be specified by listing the first variable, a colon, and then the last variable. To specify multiple variables, a single variable or a list, separate each by a comma, then invoke the R combine or c function. For example, if the list of variables in the input correlation matrix is from m02 through m05, and the variable Anxiety, then define the list in the corReorder function call according to vars=c(m02:m05,Anxiety).

Or, define the ordering with a hierarchical cluster analysis from the base R function hclust(). The same default type of "complete" is provided, though this can be changed with the parameter hclust_type according to hclust. Default input is a correlation matrix, converted to a matrix of dissimilarities by subtracting each element from 1.

Or, use the Hunter (1973) chain method. Define the ordering of the variables according to the following algorithm. If no variable list is specified then the variables are re-ordered such that the first variable is that which has the largest sum of squared correlations of all the variables, then the variable that has the highest correlation with the first variable, and so forth.

In the absence of a variable list, the first variable in the re-ordered matrix can be specified with the chain_first option.

Author(s)

David W. Gerbing (Portland State University; [e-mail])

References

Hunter, J.E. (1973), Methods of reordering the correlation matrix to facilitate visual inspection and preliminary cluster analysis, Journal of Educational Measurement, 10, p51-61.

See Also

Correlation, hclust.

Examples

# input correlation matrix of perfect two-factor model
# Factor Pattern for each Factor: 0.8, 0.6, 0.4
# Factor-Factor correlation: 0.3
mycor <- matrix(nrow=6, ncol=6, byrow=TRUE,
c(1.000,0.480,0.320,0.192,0.144,0.096,
  0.480,1.000,0.240,0.144,0.108,0.072,
  0.320,0.240,1.000,0.096,0.072,0.048,
  0.192,0.144,0.096,1.000,0.480,0.320,
  0.144,0.108,0.072,0.480,1.000,0.240,
  0.096,0.072,0.048,0.320,0.240,1.000))
colnames(mycor) <- c("V1", "V2", "V3", "V4", "V5", "V6")
rownames(mycor) <- colnames(mycor)

# leave only the 3 indicators of the second factor
#  in reverse order
#replace original mycor
mycor <- corReorder(vars=c(V6,V5,V4))

#  reorder according to results of a hierarchical cluster analysis
mynewcor <- corReorder()

#  get cluster membership for two clusters
# specify each parameter
mynewcor <- corReorder(mycor, order="hclust", n_clusters=2)

#  reorder with first variable with largest sums of squares
mynewcor <- corReorder(order="chain")

# reorder the variables according to the ordering algorithm
#  with the 4th variable listed first
# no heat map
mynewcor <- corReorder(chain_first=2, heat_map=FALSE)

mycor <- matrix(nrow=6, ncol=6, byrow=TRUE,
c(1.000,0.480,0.320,0.192,0.144,0.096,
  0.480,1.000,0.240,0.144,0.108,0.072,
  0.320,0.240,1.000,0.096,0.072,0.048,
  0.192,0.144,0.096,1.000,0.480,0.320,
  0.144,0.108,0.072,0.480,1.000,0.240,
  0.096,0.072,0.048,0.320,0.240,1.000))
colnames(mycor) <- c("V1", "V2", "V3", "V4", "V5", "V6")
rownames(mycor) <- colnames(mycor)

# can also re=order with index position of each variable
mycor <- corReorder(vars=c(4,5,6,1,2,3))