8

I have the following plot:

enter image description here

Generated with this code:

library("GGally")
data(iris)
ggpairs(iris[, 1:4], lower=list(continuous="smooth", params=c(colour="blue")),
  diag=list(continuous="bar", params=c(colour="blue")), 
  upper=list(params=list(corSize=6)), axisLabels='show')

My questions are:

  1. How can I change the correlation line to be red, now it's black.
  2. And the correlation line is buried under the scatter plot. I want to put it on top. How can I do that?
Rorschach
  • 31,301
  • 5
  • 78
  • 129
neversaint
  • 60,904
  • 137
  • 310
  • 477
  • You can control the transparency of the dots in the scatter plot by adding `alpha=0.3` to your lower list params. This will help focus the smooth lines more. – nehiljain Jun 16 '15 at 03:55
  • @nehiljain: No that won't do. When the scatter plot is dense the line will still be buried. I was thinking in [this line](http://stackoverflow.com/questions/15214489/how-to-unearth-the-buried-regression-line-in-ggplot) . But don't know how to implement it in ggpairs. – neversaint Jun 16 '15 at 04:08

2 Answers2

14

Check your version of GGally by packageVersion("GGally") and upgrade your GGally to version 1.0.1

library("GGally")
library("ggplot2")
data(iris)

lowerFn <- function(data, mapping, method = "lm", ...) {
  p <- ggplot(data = data, mapping = mapping) +
    geom_point(colour = "blue") +
    geom_smooth(method = method, color = "red", ...)
  p
}

ggpairs(
  iris[, 1:4], lower = list(continuous = wrap(lowerFn, method = "lm")),
  diag = list(continuous = wrap("barDiag", colour = "blue")),
  upper = list(continuous = wrap("cor", size = 10))
)

enter image description here

WCC
  • 1,922
  • 1
  • 14
  • 7
3

I hope there is an easier way to do this, but this is a sort of brute force approach. It does give you flexibility to easily customize the plots further however. The main point is using putPlot to put a ggplot2 plot into the figure.

library(ggplot2)

## First create combinations of variables and extract those for the lower matrix
cols <- expand.grid(names(iris)[1:4], names(iris)[1:3])    
cols <- cols[c(2:4, 7:8, 12),]  # indices will be in column major order

## These parameters are applied to each plot we create
pars <- list(geom_point(alpha=0.8, color="blue"),              
             geom_smooth(method="lm", color="red", lwd=1.1))

## Create the plots (dont need the lower plots in the ggpairs call)
plots <- apply(cols, 1, function(cols)                    
    ggplot(iris[,cols], aes_string(x=cols[2], y=cols[1])) + pars)
gg <- ggpairs(iris[, 1:4],
              diag=list(continuous="bar", params=c(colour="blue")), 
              upper=list(params=list(corSize=6)), axisLabels='show')

## Now add the new plots to the figure using putPlot
colFromRight <- c(2:4, 3:4, 4)                                    
colFromLeft <- rep(c(1, 2, 3), times=c(3,2,1))
for (i in seq_along(plots)) 
    gg <- putPlot(gg, plots[[i]], colFromRight[i], colFromLeft[i])
gg

enter image description here

## If you want the slope of your lines to correspond to the 
## correlation, you can scale your variables
scaled <- as.data.frame(scale(iris[,1:4]))
fit <- lm(Sepal.Length ~ Sepal.Width, data=scaled)
coef(fit)[2]
# Sepal.Length 
#  -0.1175698 

## This corresponds to Sepal.Length ~ Sepal.Width upper panel

Edit

To generalize to a function that takes any column indices and makes the same plot

## colInds is indices of columns in data.frame
.ggpairs <- function(colInds, data=iris) {
    n <- length(colInds)
    cols <- expand.grid(names(data)[colInds], names(data)[colInds])
    cInds <- unlist(mapply(function(a, b, c) a*n+b:c, 0:max(0,n-2), 2:n, rep(n, n-1)))
    cols <- cols[cInds,]  # indices will be in column major order

    ## These parameters are applied to each plot we create
    pars <- list(geom_point(alpha=0.8, color="blue"),              
                 geom_smooth(method="lm", color="red", lwd=1.1))

    ## Create the plots (dont need the lower plots in the ggpairs call)
    plots <- apply(cols, 1, function(cols)                    
        ggplot(data[,cols], aes_string(x=cols[2], y=cols[1])) + pars)
    gg <- ggpairs(data[, colInds],
                  diag=list(continuous="bar", params=c(colour="blue")), 
                  upper=list(params=list(corSize=6)), axisLabels='show')

    rowFromTop <- unlist(mapply(`:`, 2:n, rep(n, n-1)))
    colFromLeft <- rep(1:(n-1), times=(n-1):1)
    for (i in seq_along(plots)) 
        gg <- putPlot(gg, plots[[i]], rowFromTop[i], colFromLeft[i])
    return( gg )
}

## Example
.ggpairs(c(1, 3))
Rorschach
  • 31,301
  • 5
  • 78
  • 129