openair (version 0.5-2)

corPlot: corrgram plot with conditioning

Description

Function to to draw and visualise correlation matrices using lattice. The primary purpose is as a tool for exploratory data analysis. Hierarchical clustering is used to group similar variables.

Usage

corPlot(mydata, type = "default", cluster = TRUE, cols = "default", r.thresh = 0.8,
                         text.col = c("black", "black"),  main = "", auto.text = TRUE)

Arguments

mydata

A data frame which should consist of some numeric columns.

type

type determines how the data are split i.e. conditioned, and then plotted. The default is will produce a single plot using the entire data. Type can be one of the built-in types as detailed in cutData e.g. "season"

cluster

Should the data be ordered according to cluster analysis. If TRUE hierarchical clustering is applied to the correlation matrices using hclust to group similar variables together. With many variables clustering can greatl

cols

Colours to be used for plotting. Options include "default", "increment", "heat", "spectral", "hue", "brewer1", "greyscale" and user defined (see openColours for more details).

r.thresh

Values of greater than r.thresh will be shown in bold type. This helps to highlight high correlations.

text.col

The colour of the text used to show the correlation values. The first value controls the colour of negative correlations and the second positive.

main

The plot title; default is no title.

auto.text

Either TRUE (default) or FALSE. If TRUE titles and axis labels will automatically try and format pollutant names and units properly e.g. by subscripting the `2' in NO2.

Details

The corPlot function plots correlation matrices. The implementation relies heavily on that shown in Sarkar (2007), with a few extensions. Correlation matrices are a very effective way of understating relationships between many variables. The corPlot shows the correlation coded in three ways: by shape (ellipses), colour and the numeric value. The ellipses can be thought of as visual representations of scatter plot. With a perfect positive correlation a line at 45 degrees positive slope is drawn. For zero correlation the shape becomes a circle. See examples below. With many different variables it can be difficult to see relationships between variables i.e. which variables tend to behave most like one another. For this reason hierarchical clustering is applied to the correlation matrices to group variables that are most similar to one another (if cluster = TRUE.) It is also possible to use the openair type option to condition the data in many flexible ways, although this may become difficult to visualise with too many panels.

References

Sarkar, D. (2007). Lattice Multivariate Data Visualization with {R}. New York: Springer. Friendly, M. (2002). Corrgrams : Exploratory displays for correlation matrices. American Statistician, 2002(4), 1-16. doi:10.1198/000313002533

See Also

taylor.diagram from the plotrix package from which some of the annotation code was used.

Examples

# load openair data if not loaded already
data(mydata)
## basic corrgram plot
corPlot(mydata)
## plot by season ... and so on
corPlot(mydata, type = "season")
## a more interesting are hydrocarbon measurements
hc <- importAURN(site = "my1", year = 2005, hc = TRUE)
## now it is possible to see the hydrocarbons that behave most
## similarly to one another
corPlot(hc)