MaplePrimes - answers and comments on Question, violinplot procedure

Isn't tipping and capping automatic?

Carl Love — Tue, 19 Feb 2013 22:04:02 Z

It seems to me that the tipping and capping will come automatically if you simply apply your procedure violinplot to a KernelDensityPlot with the ranging options. In particular, to enforce positivity, in the below example I use the left= 0 option.

violinplot:= proc()
   local P;
   P:= Statistics:-KernelDensityPlot(args);
   plots:-display(
      plottools:-rotate(
         plots:-display(P, plottools:-reflect(P, [[0,0], [1,0]]))
        ,Pi/2
      )
   )
end proc:

violinplot([45,7,4,345,8,456,3,2,45,444,111,34], left= 0, axes= boxed);

Isn't this plot tipped and capped as you want?

range is better and bee swarm plot

Christopher2222 — Wed, 20 Feb 2013 03:32:17 Z

Actually using left and right distorts the distribution too much. Using range makes it better. You were right with the caps. The whiskers for the tips, yes I would want to reach min and max points without going over, however I think that offsets the data too much.

with range

Christopher2222 — Wed, 20 Feb 2013 06:09:12 Z

Using range we need to close the ends. Here's the updated proc for that. Sorry I haven't elaborated on the bee swarm plot but basically I think that is data jittered inside a violin plot. Also using method=exact speeds up the production of the plot.

Download violinplot_proc_4.mw

**edit** adding bandwidth=nops(data)*2 gives a little more flavour to the plot and a better indication of where the data points are.

capped yes

Christopher2222 — Tue, 19 Feb 2013 23:06:18 Z

Yes, okay adding the left=min(data),right=max(data) is what I wanted for capping.

Regarding tipped, what I mean is shrinking the scale of the kerneldensityplot (no specified range) so that the plot ends on the upper and lower bounds of the data.

Placement of cusps

Carl Love — Wed, 20 Feb 2013 00:37:38 Z

@Christopher2222 Ah, I guess by "caps" you meant what are called "whiskers" in a box-and-whiskers plot, the central vertical lines? Those are what I thought you meant by "tips". And I thought that by "caps" you meant what appears as horizontal lines in my plot.

So now I am guessing that you want the violin to have a cusp at the top and bottom, as in your original plot, with the top cusp at the highest data point and the bottom cusp at the lowest data point. Is that right? I believe we can achieve this with plottools:-homothety. And I guess that you want the scale of the vertical axis to stay the same. Is that correct?