how to graph pie like this?

[Fred Lee]

As above, how to draw the concentric circles?

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte teamSize
 3
 2
 3
10
 3
 3
 1
 2
 2
 8
 3
 3
 2
 4
 3
 3
 3
 5
 3
 1
 3
 5
 3
 6
 3
 2
 2
 2
 3
 3
 1
10
 3
 1
 6
 1
 2
 5
 4
 2
11
 5
 4
 1
 3
 3
 4
 4
 3
 6
 3
 2
 3
 3
 3
 2
 4
 3
 2
 1
 1
 4
 6
 3
 3
 2
 5
 3
 3
 1
 3
 3
 2
 5
 4
 2
 4
 7
 2
 3
 2
 3
 2
 2
 3
 3
 4
 4
 4
 2
 2
 2
 2
 3
 3
 2
 3
 1
 2
 1
end
graph pie ,over(teamSize)  legend(ring(0) pos(3) col(1)) pie(3,explode) plabel(1 percent) plabel(2 percent) plabel(3 percent) plabel(4 percent) plabel(5 percent)

[Maarten Buis]

That is not a good graph. Think about how we humans decode the information from a graph. The "bending" of the bars only makes it harder to compare them, especially when we compare graphs on the inside, where the bending has to be more extreme, and the outside. So just forget about that graph.

Instead look at something like this:

graph dot (percent) , over(teamSize)

[Fred Lee]

Alright, thanks.

[Nick Cox]

I agree strongly with Maarten Buis . As it happens, the graph design in #1 is not easily available in Stata without customised programming, unless there is a community-contributed command lurking somewhere, but minimally you have to show about 10 frequencies or percents and the strain on the reader would be great.

I don't understand the popularity of such designs one bit. Some bad examples appear in a recent book: see my review (and compare the others...) at https://www.amazon.com/MakeoverMonda.../dp/1119510775

Thanks for your data example. Here is a graph for the pie chart code in #1 (there could be differences if people use a different scheme) and one of several possible alternatives in Maarten's spirit.

set scheme s1color

graph pie ,over(teamSize)  legend(ring(0) pos(3) col(1)) pie(3,explode) ///
plabel(1 percent) plabel(2 percent) plabel(3 percent) plabel(4 percent) plabel(5 percent) name(Fred1, replace)

histogram teamSize, discrete percent yla(1/11, ang(h) noticks) horizontal ///
bfcolor(eltgreen*0.2) blcolor(eltgreen*2) addlabels addlabopts(mlabsize(medlarge))    ///
barw(0.9) ysc(reverse)xsc(off) ytitle("") subtitle(% of each team size) ysc(r(. 12)) name(Fred2, replace)

I often like graph-table hybrids. If the frequencies or percents are shown as text, then the corresponding axis labels are no longer needed. The histogram is helpful here to respect the detail that there were no teams of size 9. Even if the fuller dataset fills in that gap, there may be others. Showing bars with widths a little less than 1 is a way to put subtle emphasis on the discreteness of the scale, but clearly not compulsory. Even more at choice are a horizontal alignment and bar colour(s).

Do people need to be persuaded that the pie chart does not work well?