Re: [Gnuplot-info] Can gnuplot create cycle plots (also called month plots)?

[Tait <gnu...@t4...>]

> Cycle plots are useful in my work, and I'd like to know if it possible to
> create cycle plots in gnuplot.
>
> Naomi Robbins discusses cycle plots in her book ...
> as well as in this article:
> http://www.perceptualedge.com/articles/guests/intro_to_cycle_plots.pdf


As far as gnuplot is concerned, this is just a data plot, like any other. I
made a quick example: http://www.plotshare.com/index.ws/plot/524664706. It
uses linespoints for the line parts, and candlesticks to draw in the
median line.

On a stylistic note, I would not use a cycle plot. A graph should clearly
and intuitively convey its message to the audience in the simplest way
possible. I've yet to see a cycle plot that didn't complicate and obscure
the message, when a simpler approach would have worked better.

The article's claims notwithstanding, a cycle plot has the same
weaknesses as the plots it purports to replace. By grouping the x-axis
according to weekday, informtaion on trend through the week is
difficult to read. (Try telling what happened in week three, for
example: you must count one, two, three points from the left side of
each grouping.) If the cycle plot had instead grouped by week
(sub-grouping by day), the reader would have the opposite problem (what
happened on Wednesdays?). Just put the independent variable on the
x-axis, and the dependent on the y. If you have two points to make,
then do it with two graphs. Don't try to squish both points awkwardly
into one graph that makes neither point clearly.

I'll illustrate using the example from your linked article. If the
intent is to show trends by day for each week: put weeks 1-8 on a common
x-axis, put the units sold for a given day on the y-axis, and stack the
graphs for each day one on top of another. I just pulled this together
quickly; it would need y-tics and some adjustment, but it might look like
http://www.plotshare.com/index.ws/plot/785484084.

If the intent is to convey relative mean and variance from day to day,
then a box plot should be used to show average, variance, and outliers
for each day, like http://www.plotshare.com/index.ws/plot/292462185.

If the purpose was to compare week one to week four, then repeat the
box plot using week number as the independent axis. Trying to do all at
once pulls the audience's attention in multiple directions.