corrgram (version 0.1)

corrgram: Draw a correlogram

Description

The corrgram function produces a graphical display of a correlation matrix, called a correlogram. The cells of the matrix can be shaded or colored to show the correlation value.

Usage

corrgram(x, order = NULL, labels, panel=panel.shade, ...,
  lower.panel=panel, upper.panel=panel, diag.panel=NULL,
text.panel=textPanel, label.pos=0.5, cex.labels=NULL, font.labels=1, 
row1attop=TRUE, gap=0)

Arguments

x

A tall data frame with one observation per row. Only numeric values are allowed.

order

Currently, anything other than NULL will cause the variables to be ordered using principal component analysis of the correlation matrix.

labels

Not used

panel

function used to plot the contents of each panel

...

additional arguments passed to methods

lower.panel, upper.panel

Separate panel functions used below/above the diagonal

diag.panel, text.panel

panel function used on the diagonal

label.pos

vertical placement of label in diagonal panels

cex.labels, font.labels

Graphics parameter for diagonal panels

row1attop

TRUE for diagonal like " ", FALSE for diagonal like " / ".

gap

Distance between panels

Value

• No value is returned. A plot is created.

Details

This function is basically a modification of the pairs.default function with the use of customized panel functions. The off-diagonal panels are specified with panel.pts, panel.pie, panel.shade, panel.ellipse. Diagonal panels are specified with panel.txt, panel.minmax. The panel functions use col.corrgram to specify the colors. Change this function if you desire to use a different palette. The function currently can re-produce some/most of the examples in the paper by Friendly, but has not been tested on other datasets. TODO: legend, partial correlations

References

Friendly, Michael. 2002. Corrgrams: Exploratory Displays for Correlation Matrices. The American Statistician, 56, 316--324. Online: http://www.math.yorku.ca/SCS/Papers/corrgram.pdf A SAS macro by Michael Friendly is at http://www.math.yorku.ca/SCS/sasmac/corrgram.html.

See Also

The plotcorr function in the ellipse package has some similarities. The corrgram function is derived from the pairs function.

Examples

# Load datasets first
data(baseball)
data(auto)

# Most of the following examples are found in the paper by Michael
# Friendly.

# Figure 2
vars2 <- c("Assists","Atbat","Errors","Hits","Homer","logSal",
           "Putouts","RBI","Runs","Walks","Years")
corrgram(baseball[,vars2], order=TRUE, 
         main="Baseball data PC2/PC1 order",
         lower.panel=panel.shade, upper.panel=panel.pie,
         text.panel=panel.txt)

# Figure 3
baseball.cor <- cor(baseball[,vars2], use='pair')
baseball.eig <- eigen(baseball.cor)$vectors[,1:2]
e1 <- baseball.eig[,1]
e2 <- baseball.eig[,2]
plot(e1,e2,col='white', xlim=range(e1,e2), ylim=range(e1,e2))
arrows(0, 0, e1, e2, cex=0.5, col="red", length=0.1)
text(e1,e2, rownames(baseball.cor), cex=0.75)

# Figure 4a
corrgram(baseball[,vars2], main="Baseball data (alphabetic order)")

# Figure 4b.  panel functions explicitly called
corrgram(baseball[,vars2], order=TRUE,
         main="Baseball data (PC order)",
         panel=panel.shade, text.panel=panel.txt)

# Figure 5
vars5 <- setdiff(colnames(baseball), c("Name","League","Team","Position"))
corrgram(baseball[, vars5], order=TRUE,
         main="Baseball data (PC order)")

# Figure 6. Arrangement is slightly different from Friendly.
vars6 <- setdiff(colnames(auto), c("Model", "Origin"))
corrgram(auto[, vars6], order=TRUE, main="Auto data (PC order)")

# Figure 7, 8, 9
# To re-create figure 7, 8 and 9, we need to re-write corrgram to
# work with a correlation matrix instead of a data matrix
# Not yet implemented

# Figure 11.  
corrgram(baseball[,vars2], order=TRUE,
         main="Baseball correlation ellipses",
         panel=panel.ellipse, text.panel=panel.txt, diag.panel=panel.minmax)

# Reverse diagonal, use points in lower part
corrgram(baseball[,vars2], order=TRUE, row1attop=FALSE,
         main="Baseball correlation ellipses",
         upper.panel=panel.ellipse, lower.panel=panel.pts, diag.panel=panel.minmax)