scatter.hist function

Draw a scatter plot with associated X and Y histograms, densities and correlation

Draw a X Y scatter plot with associated X and Y histograms with estimated densities. Will also draw density plots by groups, as well as distribution ellipses by group. Partly a demonstration of the use of layout. Also includes lowess smooth or linear model slope, as well as correlation and Mahalanobis distances

scatterHist(x,y=NULL,smooth=TRUE,ab=FALSE, correl=TRUE,data=NULL, density=TRUE,means=TRUE, 
   ellipse=TRUE,digits=2,method="pearson",cex.cor=1,cex.point=1,
   title="Scatter plot + density",
   xlab=NULL,ylab=NULL,smoother=FALSE,nrpoints=0,xlab.hist=NULL,ylab.hist=NULL,grid=FALSE,
   xlim=NULL,ylim=NULL,x.breaks=11,y.breaks=11,
   x.space=0,y.space=0,freq=TRUE,x.axes=TRUE,y.axes=TRUE,size=c(1,2),
   col=c("blue","red","black"),legend=NULL,transparency=.5,pch=21, show.d=TRUE,
   x.arrow=NULL,y.arrow=NULL,d.arrow=FALSE,cex.arrow=1,
      line.col="blue",alpha = .05,ci=TRUE, ci.col="light blue",...)
 
scatter.hist(x,y=NULL,smooth=TRUE,ab=FALSE, correl=TRUE,data=NULL,density=TRUE,
  means=TRUE, ellipse=TRUE,digits=2,method="pearson",cex.cor=1,cex.point=1,
  title="Scatter plot + density",
  xlab=NULL,ylab=NULL,smoother=FALSE,nrpoints=0,xlab.hist=NULL,ylab.hist=NULL,grid=FALSE,
  xlim=NULL,ylim=NULL,x.breaks=11,y.breaks=11,
  x.space=0,y.space=0,freq=TRUE,x.axes=TRUE,y.axes=TRUE,size=c(1,2),
  col=c("blue","red","black"),legend=NULL,transparency=.5,pch=21, show.d=TRUE,
   x.arrow=NULL,y.arrow=NULL,d.arrow=FALSE,cex.arrow=1,
      line.col="blue",alpha = .05,ci=TRUE, ci.col="light blue",...)

Arguments

• x: The X vector, or the first column of a data.frame or matrix. Can be specified using formula input.
• y: The Y vector, of if X is a data.frame or matrix, the second column of X
• smooth: if TRUE, then add a loess smooth to the plot
• ab: if TRUE, then show the best fitting linear fit
• correl: TRUE: Show the correlation
• data: if using formula input, the data must be specified
• density: TRUE: Show the estimated densities
• means: TRUE: If TRUE, show the means for the distributions.
• ellipse: TRUE: draw 1 and 2 sigma ellipses and smooth
• digits: How many digits to use if showing the correlation
• method: Which method to use for correlation ("pearson","spearman","kendall") defaults to "pearson"
• smoother: if TRUE, use smoothScatter instead of plot. Nice for large N.
• nrpoints: If using smoothScatter, show nrpoints as dots. Defaults to 0
• grid: If TRUE, show a grid for the scatter plot.
• cex.cor: Adjustment for the size of the correlation
• cex.point: Adjustment for the size of the data points
• xlab: Label for the x axis
• ylab: Label for the y axis
• xlim: Allow specification for limits of x axis, although this seems to just work for the scatter plots.
• ylim: Allow specification for limits of y axis
• x.breaks: Number of breaks to suggest to the x axis histogram.
• y.breaks: Number of breaks to suggest to the y axis histogram.
• x.space: space between bars
• y.space: Space between y bars
• freq: Show frequency counts, otherwise show density counts
• x.axes: Show the x axis for the x histogram
• y.axes: Show the y axis for the y histogram
• size: The sizes of the ellipses (in sd units). Defaults to 1,2
• col: Colors to use when showing groups
• transparency: Amount of transparency in the density plots
• legend: Where to put a legend c("topleft","topright","top","left","right")
• pch: Base plot character (each group is one more)
• xlab.hist: Not currently available
• ylab.hist: Label for y axis histogram. Not currently available.
• title: An optional title
• show.d: If TRUE, show the distances between the groups
• d.arrow: If TRUE, draw an arrow between the two centroids
• x.arrow: optional lable for the arrow connecting the two groups for the x axis
• y.arrow: optional lable for the arrow connecting the two groups for the y axis
• cex.arrow: cex control for the label size of the arrows.
• line.col: color of the lowess or lm fit line
• alpha: When drawing confidence intervals, what is the margin of error
• ci: Draw confidence intervals
• ci.col: color of confidence intervals
• ...: Other parameters for graphics

Details

Just a straightforward application of layout and barplot, with some tricks taken from pairs.panels. The various options allow for correlation ellipses (1 and 2 sigma from the mean), lowess smooths, linear fits, density curves on the histograms, and the value of the correlation. ellipse = TRUE implies smooth = TRUE. The grid option provides a background grid to the scatterplot.

If using grouping variables, will draw ellipses (defaults to 1 sd) around each centroid. This is useful when demonstrating Mahalanobis distances.

Formula input allows specification of grouping variables as well. )

For plotting data for two groups, Mahalobnis differences between the groups may be shown by drawing an arrow between the two centroids. This is a bit messy and it is useful to use pch="." in this case.

Author(s)

William Revelle

Note

Originally adapted from Addicted to R example 78. Modified following some nice suggestions from Jared Smith. Substantial revisions in 2021 to allow for a clearer demonstration of group differences.

See Also

pairs.panels for multiple plots, multi.hist for multiple histograms and histBy for single variables with multiple groups. Perhaps the best example is found in the psychTools::GERAS data set.

Examples

data(sat.act)
with(sat.act,scatterHist(SATV,SATQ))
scatterHist(SATV ~ SATQ,data=sat.act)  #formula input

#or for something a bit more splashy
scatter.hist(sat.act[5:6],pch=(19+sat.act$gender),col=c("blue","red")[sat.act$gender],grid=TRUE)
#better yet
scatterHist(SATV ~ SATQ + gender,data=sat.act) #formula input with a grouping variable

#If using a factor for grouping, we must first convert it to numeric
iris.1 <- char2numeric(iris,flag=FALSE)
scatterHist(Sepal.Width ~ Petal.Length + Species,data=iris.1, show.d=FALSE,
   main="Fisher's Iris example") #see pairs.panels