How to obtain the lengths of semi axes of an ellipse? in R

Question

I have this set of x and y coordinates:

x<-c(1.798805,2.402390,2.000000,3.000000,1.000000)
y<-c(0.3130147,0.4739707,0.2000000,0.8000000,0.1000000)
as.matrix(cbind(x,y))->d

and I want to calculate the ellipsoid that contains this set of points, I use the function ellipsoidhull() in the package "cluster", and I get:

> ellipsoidhull(d)
'ellipsoid' in 2 dimensions:`
 center = ( 2.00108 0.36696 ); squared ave.radius d^2 =  2`
and shape matrix =
x 0.66590 0.233106
y 0.23311 0.095482
  hence, area  =  0.60406

However it's not obvious to me how I can get from these results, the lengths of the semi-major axes of this ellipse.

Any idea?

Thank you very much in advance.

Tina.

Answer 1

The square of the semi-axes are the eigenvalues of the shape matrix, times the average squared radius.

x <- c(1.798805,2.402390,2.000000,3.000000,1.000000)
y <- c(0.3130147,0.4739707,0.2000000,0.8000000,0.1000000)
d <- cbind( x, y )
library(cluster)
r <- ellipsoidhull(d)
plot( x, y, asp=1, xlim=c(0,4) )
lines( predict(r) )
e <- sqrt(eigen(r$cov)$values)
a <- sqrt(r$d2) * e[1]  # semi-major axis
b <- sqrt(r$d2) * e[2]  # semi-minor axis
theta <- seq(0, 2*pi, length=200)
lines( r$loc[1] + a * cos(theta), r$loc[2] + a * sin(theta) )
lines( r$loc[1] + b * cos(theta), r$loc[2] + b * sin(theta) )

Answer 2

You can do this:

exy <- predict(ellipsoidhull(d)) ## the ellipsoid boundary
me <- colMeans((exy))            ## center of the ellipse

Then you compute the minimum and maximum distance to get respectively minor and major axis:

dist2center <- sqrt(rowSums((t(t(exy)-me))^2))
max(dist2center)     ## major axis
[1] 1.264351
> min(dist2center)   ## minor axis
[1] 0.1537401

EDIT plot the ellipse with the axis:

plot(exy,type='l',asp=1)
points(d,col='blue')
points(me,col='red')
lines(rbind(me,exy[dist2center == min(dist2center),]))
lines(exy[dist2center == max(dist2center),])