R help - Sep 2006 - Adding percentage to Pie Charts (was (no subject))

Have you read the books by Cleveland?

His experiments show that most people do better estimating things and
comparing things on a linear scale rather than looking at angles and
areas (also see
http://biostat.mc.vanderbilt.edu/twiki/pub/Main/StatGraphCourse/graphsco
urse.pdf)

With a dot chart you can set the axis to go from 0 to the total of all
groups (see the example I sent before, it could have had the numbers on
the x-axis, but still included the total), that means that points near
the middle of the line represent about 50%, looking at how close the
point is to the left lets you estimate the percentage and most people
(you may differ) do a better job of estimating that percentage from the
position of the dot than from an angle or area.  If you feel the need to
specify the percentages along the side of a dot chart, then at least
they are lined up vertically for easy comparison (my example would have
been better if a lot of the vertical space had been removed so the
pieces of interest were closer together), the pie chart would generally
have the percentages non-aligned causing more work for the viewer to
compare them.

Dotcharts also remove the dependence/temptation to use color and any
psycolocical influences that may have on the interpretation.

I have yet to see a pie chart that was better at conveying the true
nature of the data than a well done dot chart of the same data, I have
seen multiple cases where the dot chart showed truths about the data
that were not apparent in the corresponding pie chart.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at intermountainmail.org
(801) 408-8111
 

-----Original Message-----
From: r-help-bounces at stat.math.ethz.ch
[mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Anupam Tyagi
Sent: Tuesday, September 19, 2006 12:52 AM
To: r-help at stat.math.ethz.ch
Subject: Re: [R] Adding percentage to Pie Charts (was (no subject))

Greg Snow <Greg.Snow <at> intermountainmail.org> writes:
> 
> You may want to rethink your whole approach here:
> 
> 1. Pie charts are usually a poor choice of graph, there are better 
> choices.
> 2. Adding percentages to a pie chart is a way of admitting that the 
> pie chart is not doing the job.
> 3. If you want people to compare percentages, then a table is what is 
> needed.
> 4. A pie chart with percentages added is essentially a colorful but 
> poorly layed out table.
> 
> Consider using a dotplot instead of a pie chart, it changes the job of
> the viewer from comparing areas/angles (done poorly by humans) to 
> comparing positions along a common scale (done well by humans).
I think dot charts (plots) are very useful, but they are not substitutes
for a pie chart: they do not show a comparison between the total and the
individual value; have a different scale (linear, usually), and are
visually not suitable to answer some questions that a pie chart can
answer (is the value approximately less than a fourth of the total? Is
it less than half?). For some of these questions, even dot-charts
require a value label, or the user doing mental calculations to guess
approximations.

I think I am quite attuned to getting approximate fractions from a
pie-chart in shorter time, than on a linear scale like the dot-chart.

A modification in a pie chart that draws overlapping areas with a common
start point at the top of the circle, can make is more informative than
a dot-chart.
Something like:
* Start drawing at the top of the circle, as zero (degree/area).
* Draw the representation of every value starting from the top, as zero,
representing it as a labled line from the center of the circle to the
boundary (can use colors where possible).
* Use two lables for the circular axis, inside one for percentages,
outside for values.

What is the simplest way to draw this in R?

Anupam.

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Greg Snow <Greg.Snow <at> intermountainmail.org> writes:
> Have you read the books by Cleveland?
I do not recall reading Cleveland's book; I have read one by Tufte. You
raise
some interesting issues there. I agree with some, I could not clearly understand
some other things you mention. 

I think visual perception is aquired, in part. So if I were presenting data to
viewers who took carpentry or other such classes in highschool I may be tempted
to use dotcarts.

An interesting experiment: have kids compare pieces of pie or bread-sticks over
a dinner, and check how they do.  They should not have taken a carpentry class.
I use dot-charts, they are useful. Sometimes pie carts are useful too, because
people are so used to using and seeing them over a long time. Ofcourse, they can
be improved.

Also, it may be possible to put points of a dot-chart on a single straight line,
label them with a pointing line, and get better perception. There is poor
perception of the horizantal distance, by having to view that extra vertical
distance in a dotchart. However, it is useful to have the vertical axis in
Lattice plots, but not in stand-alone dot-charts.

Anupam.