How can a data ellipse be superimposed on a ggplot2 scatterplot?

Question

I have an R function which produces 95% confidence ellipses for scatterplots. The output looks like this, having a default of 50 points for each ellipse (50 rows):

           [,1]         [,2]
 [1,]  0.097733810  0.044957994
 [2,]  0.084433494  0.050337990
 [3,]  0.069746783  0.054891438

I would like to superimpose a number of such ellipses for each level of a factor called 'site' on a ggplot2 scatterplot, produced from this command:

> plat1 <- ggplot(mapping=aes(shape=site, size=geom), shape=factor(site)); plat1 + geom_point(aes(x=PC1.1,y=PC2.1))

This is run on a dataset, called dflat which looks like this:

site      geom         PC1.1        PC2.1       PC3.1        PC1.2       PC2.2
1 Buhlen 1259.5649 -0.0387975838 -0.022889782  0.01355317  0.008705276  0.02441577
2 Buhlen  653.6607 -0.0009398704 -0.013076251  0.02898955 -0.001345149  0.03133990

The result is fine, but when I try to add the ellipse (let's say for this one site, called "Buhlen"):

> plat1 + geom_point(aes(x=PC1.1,y=PC2.1)) + geom_path(data=subset(dflat, site="Buhlen"),mapping=aes(x=ELLI(PC1.1,PC2.1)[,1],y=ELLI(PC1.1,PC2.1)[,2]))

I get an error message: "Error in data.frame(x = c(0.0977338099339815, 0.0844334944904515, 0.0697467834016782, : arguments imply differing number of rows: 50, 211

I've managed to fix this in the past, but I cannot remember how. It seems that geom_path is relying on the same points rather than plotting new ones. Any help would be appreciated.

Solution 1

Maybe this could help you:

#bootstrap
set.seed(101)
n <- 1000
x <- rnorm(n, mean=2)
y <- 1.5 + 0.4*x + rnorm(n)
df <- data.frame(x=x, y=y, group="A")
x <- rnorm(n, mean=2)
y <- 1.5*x + 0.4 + rnorm(n)
df <- rbind(df, data.frame(x=x, y=y, group="B"))

#calculating ellipses
library(ellipse)
df_ell <- data.frame()
for(g in levels(df$group)){
df_ell <- rbind(df_ell, cbind(as.data.frame(with(df[df$group==g,], ellipse(cor(x, y), 
                                         scale=c(sd(x),sd(y)), 
                                         centre=c(mean(x),mean(y))))),group=g))
}
#drawing
library(ggplot2)
p <- ggplot(data=df, aes(x=x, y=y,colour=group)) + geom_point(size=1.5, alpha=.6) +
  geom_path(data=df_ell, aes(x=x, y=y,colour=group), size=1, linetype=2)

Output looks like this:

Here is more complex example.

Solution 2

Keelan Evanini, Ingrid Rosenfelder and Josef Fruehwald ([email protected]) have created a ggplot2 stat implementation of a 95% confidence interval ellipses (and an easier way to plot ellipses in ggplot2):

GitHub stat-ellipse.R

their site

You can use it as:

library(ggplot2)
library(devtools)
library(digest)
source_url("https://raw.github.com/low-decarie/FAAV/master/r/stat-ellipse.R")    
qplot(data=df, x=x, y=y, colour=colour)+stat_ellipse()

To create the data

set.seed(101)
n <- 1000
x <- rnorm(n, mean=2)
y <- 1.5 + 0.4*x + rnorm(n)
colour <- sample(c("first", "second"), size=n, replace=T)
df <- data.frame(x=x, y=y, colour=colour)

Comments

[Etienne Low-Décarie]

It appears the new version of ggplot2 broke stat_ellipse as it gradually deprecates certain other approaches in favour of S3.

[Etienne Low-Décarie]

I have create a fix which is now found in the answer.

[sautedman]

Odd behavior can occur if color is turned off. Specifically without color=... in the plot call, there is a line drawn between the edges of the ellipses. This can be avoided with group=group (using the infelicitous variable name).

[EA304GT]

stat_ellipse() seems to be working just fine with ggplot2 3.2.0 (latest in CRAN at this moment.)