Boxplots over two categorical variables - Adjust box gap and colour

[Guillaume Geri]

Hi Statalist

I use Stata 14.1 on MacOS. I'm stuck with a (probably) simple thing.
A sample of my dataset can be obtained using the following code:

* Example generated by -dataex-. To install: ssc install dataex
clear
input float std_PHYS byte insee float(id_controle class)
  50.6022 0 1 2
 55.27794 1 1 1
 57.39201 1 1 1
33.684685 1 1 1
  47.8986 1 1 1
 26.17278 1 1 1
 48.88623 1 1 1
 48.38394 1 1 1
 54.13319 1 1 1
 47.31923 1 1 1
23.920927 1 1 1
 44.27946 1 1 1
 55.44976 1 1 1
 42.15575 1 1 1
 51.50533 1 1 1
 55.01159 1 1 1
 55.18439 0 2 4
65.090256 1 2 3
 56.41214 1 2 3
 23.55465 1 3 5
end

I want to graph (using graph box) a continuous variable according to two categorical variables. My final goal is to get a graph box like this: large box gap between categories and adjusting colour across the second category.

But, when I use two over() categories, I get gaps as I want but I cannot figure out how to personalize colours of boxes 2,4 and 6.

graph box std_PHYS, over(insee) over(id_controle)  asyvar

And when I use one combining variable as follows

graph box std_PHYS, over(class)  asyvar box(1, color(red)) box(2, color(navy)) box(3, color(red)) box(5, color(red)) box(4, color(blue)) box(6, color(green))

, I'm not able to adjust two different gaps:

As shown in the first graph (done with Graph Editor), I guess it's possible using code but I did not figure out how to do that...

Any idea?
Thank you very much

[Andrew Musau]

So you have the composite variable "class" and all you appear to want is a plot over the variable "id_controle". The first graph should guide you on what to do here. Just include the option -nofill- to omit empty categories that inevitably exist in your desired graph.

graph box std_PHYS, over(class) over(id_controle) asyvar nofill box(1, color(red)) box(2, color(navy)) box(3, color(red)) box(5, color(red)) box(4, color(blue)) box(6, color(green))

[Guillaume Geri]

Thank you very much Andrew
it works perfectly
Best