Robert Lundqvist wrote:
>
> I am not sure as I have not really tried it, but as far as I can see,
> there should be possibilities to overcome at least some obstacles by
> "audio-lising" rather than "visualising". Using sound(s), i e signals at
> different frequencies with more or less noise could make it possible to
> compare samples. Anyone who have tried? In case it works, guess it should
> be useful in both statistics teaching and stats practice for those who can
> see too.
If you don't use the time dimension to separate the frequencies, you'd
need to be an orchestral conductor to keep it all straight.
Volume is probably not a good dimension - I don't think people can
distinguish very many absolute volumes. As evidence, note that musicians
have found the need for only about eight absolute volume indications:
(ppp,pp,p,mp,mf,f,ff,fff) whereas 88 or more distinct pitches are
scored.
That leaves us with melody (pitch as a function of time); not bad for a
time series perhaps, usable for a tally, and naturally adapted to a log
transformation (taking successive C's as 1, 10, 100,... units, and C,
D, Eb, E, F, G, A, Bb, B as 1,2,3,4,5,6,7,8,9 ?) One could argue for
the use of Ab rather than the harmonically preferable Eb, as it creates
longer intervals at the lower end of the scale. The E-F semitone is a
flaw, but it avoids the discordant tritone F# on the tonic, which would
tend to throw the ear off. I think keeping a reasonably clear tonic note
would be essential, ruling out either chromatic scales (C=1, C#=1.5,
D=2, D#=2.5, E=3, F=3.5, F#=4, G=5, G#=6, A=7, A#=8, B=9) or wholetone
scales (C=[1-2), D=[2,3), E=[3,4) F#=[4,6), G#=[6,8) A#=[8,10).)
Otherwise these would be efficient. One could argue for the pentatonic
scale, though crude, as widely recognizable. (Nonmusicians: sing the
first line "Starry, starry night" of Don Maclean's "Vincent". That's the
pentatonic scale. Or play the black keys on a piano.)
Chords for paired data? Too painful.
Composers such as Bach used to use the letter names or solmisation
syllables (do,re,mi...) of notes to encode names or mottos. With a
system like this one cound have a "Fugue on the Iris Data Set." The
mind boggles.
Overall: interesting idea, but probably not the first one I'd try. I'd
go for touch, as Louis Braille did.
I would think that histograms, boxplots, scatterplots, and dotplots
could be converted to a tactile computer monitor very easily, embossed
onto appropriate paper with a stylus and template (as in handwritten
Braille), drawn with one of those raised-ink pens that are now being
sold as novelties, or printed, in true Tukey EDA style, with an
unmodified Braille printer. [The dotplot would be slightly more
efficient as you could get 2 columns of up to 3 symbols in a single
character. Note also that all Braille is monospaced, ideal for
character graphics.]
-Robert Dawson
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
