[Marc Kaulisch]

I believe the addplot() option really only works well with plots that do not contain subgraphs. This is a limitation I cannot change, I'm afraid. You could, however, apply the addplot command, with which you can address individual subgraphs (see the help file).

Cool, you have made it possible with the new release. I have tested it and it works.

[Ben Jann]

As Marc noticed, I fixed the issue with addplot() in combination with by(). violinplot now makes sure that the correct observations are available to addplot() within each subgraph. Type

. ssc install violinplot, replace

to obtain the update.

Some further additions:

• it is now possible to display the IQR box as a shaded area or as lines across the density, rather than as a bar; likewise, median and mean can now be displayed as lines (see example below)
• the statistics used for box, median and mean can now be customized; for example, type box(stat(p10 p90)) to use the 10% and 90% quantiles for the box, rather than the default 25% and 75% quantiles
• it is now possible to sort results across over groups; see subpption sort() in over()

Example: Box as shading, median as line

. sysuse nlsw88
. violinplot wage if grade>10, over(grade) vertical box(type(fill)) median(type(line))

[Ben Jann]

A further update to violinplot is available from SSC. The install the update type

. ssc install violinplot, replase
. ssc install dstat, replace
. ssc install moremata, replace

(A small change was necessary in dstat and moremata, on which violinplot depends; this is why these packages also need to be updated.)

The most important change is that violinplot now has a split() option to create half violins. Example:

. sysuse nlsw88
. violinplot wage ttl_exp tenure, pdf(ll(0)) split(union) key(fill)

ben

[Ben Jann]

Yet more features have been added to violinplot. To install the new version, type

ssc install violinplot, replace

For a list of changes see https://github.com/benjann/violinplot. For example, it is now possible to print only half of a violin (options left and right) and it is also possible to include the data in the plot, see option rag(). Here are some examples:

sysuse nlsw88
violinplot wage, left rag(offset(-.005) mlcolor(%20)) dscale(.)

sysuse auto
violinplot price, over(foreign) vertical rag(spread(5) msymbol(o) mcolor(maroon%50) mlcolor(%0))

[Kuber Bathla]

Hi Ben! Thanks a lot for all the information and work on vioplot in STATA. I have been trying to use it myself and have a dataset with variables consumption (float), paddy (float, with values 0 and 1) and round (float) with values ranging from 1-19. But somehow the over option does not seem to work with either paddy or round. The code I am trying to use:

violinplot total_consumption_per_day, pdf(ll(0)) noline fill(fcolor(%50) lcolor(%100)) over(paddy)

Can you please tell me what the problem could be here?

[Kuber Bathla]

Hi Ben! Thanks a lot for all the information and work on vioplot in STATA. I have been trying to use it myself and have a dataset with variables consumption (float), paddy (float, with values 0 and 1) and round (float) with values ranging from 1-19. But somehow the over option does not seem to work with either paddy or round. The code I am trying to use:

violinplot total_consumption_per_day, pdf(ll(0)) noline fill(fcolor(%50) lcolor(%100)) over(paddy)

Can you please tell me what the problem could be here?

For context, I am using STATA 15

[Ben Jann]

Need more info. What do you mean by "does not seem to work"? Is there an error message? If not, in what way is the result different from what you expected?

[Kuber Bathla]

It simply says "invalid syntax". I think it has something to do with the over() function, since the command works fine if I remove it. If I use the following code:

violinplot total_consumption_per_day, pdf(ll(0)) noline fill(fcolor(%50) lcolor(%100))

I get the following graph:

But ideally I want the plot to have two categories, just like you have done with the union variable in your example. Also, I tried using the sample dataset nlsw88, but even then I get the same error message if I try to use the over() function. Works fine if I don't use it. Any idea what could be the issue? Thanks for your time.

[Nick Cox]

set trace on 
set tracedepth 1

before running the command to see where it fails. You may need to bump up the trace depth.

Detail: over() is an option. not a function.

Quite different:
What underlies the data here? Are there lots of zeros? It would be interesting (to me) to get a copy or a tabulation of the data to be able to compare with say a quantile plot.

[Ignacio Herrando]

Hi! Need help with plotting violin plots...
I have been modifying this command:

violinplot Difference_CH2tau1, split(MMRStatus) noline fill(fintensity(30)) mean(type(m) m(o) mc(white)) box(type(bar)) med(type(l)) rag(spread(30) msymbol(o) mcolor(%60) mlc(%0))

To obtain the following plot...

But when I try to rescale the density function, the box plot and the density is separated, like this:

Is there any way to rescale the density plot in order to obtain more uniform plots?

Thank you in advance for your help.

By the way, the new violin plot command is awesome!

[Nick Cox]

#25

Please show the results of

dataex  Difference_CH2tau1  MMRStatus)

to allow people to experiment. https://www.statalist.org/forums/help#stata explains why and how.

[Ignacio Herrando]

My apologises, I have been without internet. Here are the data...

* Example generated by -dataex-. For more info, type help dataex
clear
input float Difference_CH2tau1 long MMRStatus
-.16197658 2
. .
-.007025041 2
-.10418536 1
.27124527 2
.1481722 2
. .
. .
-.07709967 1
. .
-.06960077 1
-.08417155 1
-.02936337 2
. .
-.18138254 1
.0463042 2
. .
-.071427256 1
-.1559932 2
. .
-.02504033 2
-.0346483 2
-.010934993 2
.07938742 2
.06219372 2
-.03885114 2
.026179945 2
.035119455 2
-.12766866 2
. .
.00698373 2
.02014844 2
-.04258387 2
. .
-.22304997 1
-.005390899 2
-.01830321 2
.02029349 2
-.014603627 2
. .
. .
.008351903 .
.0799145 2
-.0823106 2
.03225925 2
-.1486648 2
-.02678861 2
-.003597375 2
.07531597 2
-.018329207 2
. .
-.13695839 2
.00992492 2
.1639622 2
. 2
-.13790636 2
. .
-.047202 2
-.07472075 2
-.1835154 2
-.13675424 2
.034751717 2
.15841062 2
end
label values MMRStatus MMRStatus
label def MMRStatus 1 "Deficient", modify
label def MMRStatus 2 "Proficient", modify
[/CODE]

[Nick Cox]

Thanks for the data example. Questions about violinplot I will happily leave to the guru, Ben Jann.

Just curious on what might work well too. I used stripplot from SSC.

stripplot Diff, over(MMR) height(0.6) centre cumul cumprob vertical box(barw(0.05)) pctile(0) boffset(-0.4) refline xla(, noticks)

I find a direct quantile or cumulative display easier to read than the jittering. And a density function estimate from 7 data points is a bit of a stretch. The longer lines show means.

The key points are banal.

Experienced people just won't agree where taste and judgment are needed.

Which design suits your data, your purposes and your intended readership? That is your call.

[Kuber Bathla]

Hi Nick and Ben. Sorry for not being able to respond. Nick Cox, I tried what you suggested, and the code seems to stop here:

- if `pstyles'!=1 {
= if 1!=1 {
  numlist "1/`=min(`pstyles', 15)'"
  local pstyles `r(numlist)'
  }
- }
- forv p = 1/`k_`PID'' {
= forv p = 1/. {
invalid syntax

But I am not sure how to interpret this. Also attaching the log file in case you wish to check the entire output. Will appreciate any help.

Also, here is an example of my dataset. The values in total_consumption_per_day represent mean consumption in terms of hours for agricultural pumpsets, measured using hour meter devices.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(total_consumption_per_day paddy)
 125.25137 0
         . 0
      18.9 0
         0 0
 37.449398 0
  76.77143 0
 19.937143 0
 16.236254 0
  41.04235 0
  82.77869 0
 141.02322 0
  4.226332 0
 131.72212 0
         0 0
 37.910713 0
 19.958385 0
 19.415533 0
 14.906833 0
  18.10388 0
  49.62073 0
 1.7585015 0
  9.646296 0
 154.54443 0
  98.43455 0
   40.5741 0
  47.96319 0
  30.27508 0
  51.62422 0
  7.257558 0
  177.9645 0
   85.9939 0
  30.81851 0
         . 0
    140.72 0
  2.504587 0
         0 0
 10.916012 0
         0 0
 117.22666 0
  3.136517 0
 3.0946524 0
 140.09811 0
 1.6666666 0
         0 0
  5.925696 0
  85.22687 0
  62.15059 0
  4.914019 0
  5.312694 0
  7.570588 0
.003680982 0
  177.8084 0
  44.24389 0
         0 0
         0 0
 18.867924 0
 111.47853 0
  50.17107 0
  .9648902 0
  59.81667 0
         0 0
 133.50095 0
 1.3446254 0
  59.55138 0
  84.46487 0
  41.34985 0
         0 0
  9.152866 0
 4.6560745 0
  44.26038 0
  2.467826 0
  43.30384 0
  28.12215 0
  24.49059 0
 157.89146 0
 120.19514 0
 25.969326 0
  24.63889 0
  21.08647 0
 2.3226993 0
  5.003704 0
 126.68902 0
         0 0
 102.83743 0
 111.19701 0
 123.88777 0
  13.34277 0
       540 0
         0 0
  7.898508 0
  48.08802 0
         0 0
  8.163935 0
  89.50286 0
 4.3003097 0
  37.20743 0
     10.95 0
  19.57612 0
 2.0337424 0
  9.210366 0
end

[Nick Cox]

#29 is really for Ben Jann.