Skip to content
Prev 96708 / 398500 Next

Plotting league tables/ caterpillar plots

Eik Vettorazzi sent the following  at 24/07/2006 13:04:
Interesting, never heard the term "caterpillar plot" but I like that.  I
also call related things "biplane plots": the body is the parameter and
the wings (not really biplanes at all are they) are the 95% CI for the
parameter.  However, these caterpillar plots are rather like forest
plots or "trees plots" (I argue that you can see both the wood and the
trees!)

I tend to plot them rotated by 90 degrees and without such good
labelling as in the Spiegelhalter paper, but this code may help you.
The "traffic lights" in this superimpose overall quartiles which is used
in some colleagues' way of looking at league tables.  I've used similar
plots for proportions, means, correlations and even alpha reliability
parameters, it's fairly easy to substitute different parameters and
their CIs.  I prefer to use a more exact estimator of the CI of the
median than the one that, if I remember rightly, is the default in the
boxplot which is, I think a much quicker but less robust estimator of
the CI for each sample.

I'd appreciate gentle constructive criticism of my coding, I know I'm a
much better psychotherapist than a programmer!

C