boxplot, notches, etc.

Hi Ben,

I like the odd looking notches, they reveal
that the data comprising the box plot need 
more careful review.

Tukey et al (1) comment that
"It should be noted that the convention has been
adopted that, should the notch lie outside either
hinge, an unnotched box, plotted with dashed lines,
is displayed for that group indicating low confidence
in it."
While this convention appears to have been abandoned,
the wierd-looking notches indicate a group with odd
behaviour warranting more careful consideration.

So I vote to retain odd-looking notches.

Et al and Tukey (1) discuss their reasoning and empirical findings
as to the formulation for the notch width.  They used
median +/- 1.7(1.25R/1.35sqrt(N)) which reduces to
median +/- 1.574*(IQR)/sqrt(N)

Perhaps similar reasoning can resolve the issue of
notches for boxplots of log(y)?
I'll see if I can work through the normal - lognormal
transformation machinery and propose a reasonable
notch strategy for the log(y) case.  Maybe someone
out there has already done so?

(1) R. McGill, J. Tukey, W. Larsen (1978) "Variations of Box Plots"
The American Statistician, Vol. 32, No. 1, pp 12-16
http://www.jstor.org/view/00031305/di020553/02p0045u/0

Cheers

Steven McKinney

Statistician
Molecular Oncology and Breast Cancer Program
British Columbia Cancer Research Centre

email: smckinney at bccrc.ca

tel: 604-675-8000 x7561

BCCRC
Molecular Oncology
675 West 10th Ave, Floor 4
Vancouver B.C. 
V5Z 1L3
Canada

-----Original Message-----
From: r-devel-bounces at r-project.org on behalf of Ben Bolker
Sent: Tue 10/17/2006 6:09 AM
To: r-devel at stat.math.ethz.ch
Subject: Re: [Rd] boxplot, notches, etc.

Ben Bolker <bolker <at> zoo.ufl.edu> writes:

boxplot, notches, etc.

Thread (3 messages)