[R] Thinking about using two y-scales on your plot?

[Martin Maechler]

>>>>> "ASR" == Allen S Rout <asr at ufl.edu>
>>>>>     on Wed, 02 Apr 2008 10:39:31 -0400 writes:

    ASR> "hadley wickham" <h.wickham at gmail.com> writes:
    >> Please read this first:
    >> http://www.perceptualedge.com/articles/visual_business_intelligence/dual-scaled_axes.pdf
    >> 
    >> It's a reasoned discussion of why it's a bad idea and proposes some
    >> alternative methods.


    ASR> I've read the article, and I'm not sure if I'm in violation. :)

    ASR> I've got a series of graphs I generate in R to illustrate backup
    ASR> activity.  These are scaled, primarily, in bytes.  But different
    ASR> activities have different costs per byte, and I augment the bytes
    ASR> scale with dollars.

    ASR> http://docs.osg.ufl.edu/tsm/current/ext/UFEXCH-MBX01.AD.UFL.EDU-all.html

    ASR> So the blue series corresponds to tbe blue dollars scale (and bytes)
    ASR> and the green and red points correspond to the green scale (and
    ASR> bytes).

so you in fact have even three scales, where the first one
(bytes) is common to all the data drawn.

    ASR> Am I being naughty? 

    ASR> - Allen S. Rout

No. You have an interesting case.
I can well imagine that people devise a better plot than yours,
but I doubt a bit, since you can always say that your primary
interest lies in the bytes (and not the costs) which makes the
plot you did very sensible.

Originally I thought  you were talking about the following
simple situation which Stephen Few's paper actually did mention:

*One* curve / points and simply two scales for the same thing.
Very valid and reasonable.

An even simpler situation would be a time series of temperature
measurements, with the first axis using international standard
degree Celsius and the second axis using the old degree
Fahrenheit which are said to be still in use in some remote 
parts of the world ;-)

Martin