Not your question, but in most situations a reduction to four box plots
is just throwing away a lot of detail that might
be interesting or useful -- and would not be confusing if
presented carefully. You might see much, much
more on other plots.
Box plots, in my not so humble opinion, are perhaps optimal
for (say) 30 or 100 groups when you desperately need severe reduction
to see the wood for the trees -- but vastly overused for comparisons
with a very few groups (to say nothing of their over-use for
single distributions).
For just four groups, dot plots, strip plots,
beam plots, quantile plots, whatever can show much, much more.
John Tukey himself (Exploratory data analysis, Addison-Wesley,
Reading, MA, 1977), who re-invented the box plot, used Rayleigh's
original data (which led to the discovery of argon and a Nobel Prize)
to show the limitations of box plots. This dataset is bimodal (which
turns out to be the key point) but a box plot utterly obscures that.
Nick
[email protected]
Thomas Masterson
> I'm running in to a problem producing the graph I want.
> Running two separate box plots and combining them gives
> me a graph with one title, but a lot of space taken up
> with category labels. Using the by() option seems like
> it should be the answer.
> Here's what I think is the right code:
>
> . graph hbox lnw [fw=wgt], over(cat) by(group, noiy) title("Title")
>
> here's what I get:
>
> Title Title
> group1 group2
> | |
> cat1 | box cat1 | box
> | |
> cat2 | box cat1 | box
> |_________ |___________
> scale scale
>
> and here's what I want it to look like:
>
> Title
> group1 group2
> | |
> cat1 | box | box
> | |
> cat2 | box | box
> |___________ |____________
> scale scale
>
> Can anybody see what I'm doing wrong?
> I also tried the -noiyl- in the -by()-
> options, but got the same output.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
