How can I make a violin plot in PGFPlots

Question

Taking the data from this question:

Spread the dots in boxplot in pgfplots

How can I make a violin plot.

Answer 1

There is a quite new CTAN package called tikzviolinplots that fills this gap seamlessly (at least for my purposes). It follows the MWE (Minimum Working Example) of this package:

\documentclass{article}

\usepackage{pgfplots}
\usepackage{tikzviolinplots}
\usepgfplotslibrary{external}
\tikzexternalize
\usepackage{minted}
\usemintedstyle{gruvbox-light}
\usepackage{scontents}

\begin{scontents}[write-out=violinandboxplotexample.dat]
A   B   C   D
0.876   0.574   2.175   1.684
1.015   0.822   2.121   2.290
0.995   1.266   1.916   2.184
1.101   0.740   2.044   2.065
1.222   0.636   2.004   2.248
0.712   1.753   2.038   2.200
1.063   0.873   1.974   1.865
0.973   1.020   2.145   1.895
1.174   0.763   2.052   2.240
1.057   1.293   2.079   2.360
1.004   0.853   2.064   1.695
0.909   0.694   2.136   1.837
1.108   0.849   2.000   1.744
1.210   0.468   2.010   1.820
0.958   0.767   1.824   2.158
1.119   0.940   2.034   1.846
1.000   0.962   1.967   2.545
0.954   1.551   1.833   1.816
0.918   0.755   1.914   2.369
1.093   0.973   1.997   1.543
\end{scontents}

\begin{document}

\begin{tikzpicture}
    \violinsetoptions[
        averages,
        data points,
        scaled,
    ]{
        xmin=0,xmax=5,
        ymin=0,ymax=3,
        xlabel style={
            yshift = {-2*height("a")}
        },
        ymajorgrids=true,
        ylabel={Same property},
    }
    \violinplotwholefile[%
        primary color=red,
        secondary color=blue,
        indexes={A,B,C,D},
        spacing=1.0,
        labels={%
            $\alpha$,
            $\beta$,
            $\gamma$,
            $\delta$,
        },
        col sep=tab,
        dataset size=1pt,
        dataset mark=*,
        dataset fill=black!50!white,
        dataset fill opacity=1.0,
        average mark=x,
        average size=5pt,
    ]{violinandboxplotexample.dat}
\end{tikzpicture}

\end{document}

Finally there is a CTAN package what address this problem! :) Thanks are due to Pedro Callil-Soares...

edit: You must use the -shell-escape option to make it work.

Answer 2

Gnuplot can do it - see https://gnuplot.sourceforge.net/demo/violinplot.html . With some tricks, it is possible to make PGFPlots do it with help of raw gnuplot. It is not possible to do it with a table as in the link, because PGFPlots insist on having points created from the first plot. -so the code saves three additional files "kdensity1.dat", "kdensity2.dat" and "kdensity3.dat"

To run the code, you need to have Gnuplot installed and you need --shell-escape.

\begin{filecontents*}{data1.dat}
5   1
6   1
3   1
3   1
7   1
8   1
3   1
8   1
7   1
9   1
5   1
\end{filecontents*}
\begin{filecontents*}{data2.dat}
12  2
14  2
12  2
8   2
22  2
22  2
12  2
12  2
7   2
12  2
2   2
\end{filecontents*}
\begin{filecontents*}{data3.dat}
20  3
20  3
21  3
20  3
22  3
21  3
20  3
24  3
22  3
21  3
20  3
\end{filecontents*}
\documentclass[tikz, border=1 cm]{standalone}
\usepackage{pgfplots}
\pgfplotsset{compat=1.18}
\usepgfplotslibrary{fillbetween}
\begin{document}
\begin{tikzpicture}
\begin{axis}[
xmin=0, xmax=25,
ymin=0, ymax=4,
ytick={1,2,3},
yticklabels={Condition 1, Condition 2, Condition 3},
]
\addplot[draw=none] gnuplot[raw gnuplot] {
plot 0; %included to avoid warning of empty plot
set table "kdensity1.dat";
plot "data1.dat" using 1:2 smooth kdensity bandwidth 1.;
unset table;
set table "kdensity2.dat";
plot "data2.dat" using 1:2 smooth kdensity bandwidth 1.;
unset table;
set table "kdensity3.dat";
plot "data3.dat" using 1:2 smooth kdensity bandwidth 1.;
unset table;
};

\addplot[draw=none, name path=p1, y filter/.expression={1+0.1*y}] table {kdensity1.dat};
\addplot[draw=none, name path=m1, y filter/.expression={1-0.1*y}] table {kdensity1.dat};
\addplot[red!50] fill between [of=p1 and m1];

\addplot[draw=none, name path=p2, y filter/.expression={2+0.1*y}] table {kdensity2.dat};
\addplot[draw=none, name path=m2, y filter/.expression={2-0.1*y}] table {kdensity2.dat};
\addplot[green!50] fill between [of=p2 and m2];

\addplot[draw=none, name path=p3, y filter/.expression={3+0.1*y}] table {kdensity3.dat};
\addplot[draw=none, name path=m3, y filter/.expression={3-0.1*y}] table {kdensity3.dat};
\addplot[blue!50] fill between [of=p3 and m3];

\end{axis}
\end{tikzpicture}
\end{document}

Edit:

Found that the extra files is not needed after all, if gnuplot is called twice for each plot.

\begin{filecontents*}{data1.dat}
5   1
6   1
3   1
3   1
7   1
8   1
3   1
8   1
7   1
9   1
5   1
\end{filecontents*}
\begin{filecontents*}{data2.dat}
12  2
14  2
12  2
8   2
22  2
22  2
12  2
12  2
7   2
12  2
2   2
\end{filecontents*}
\begin{filecontents*}{data3.dat}
20  3
20  3
21  3
20  3
22  3
21  3
20  3
24  3
22  3
21  3
20  3
\end{filecontents*}
\documentclass[tikz, border=1 cm]{standalone}
\usepackage{pgfplots}
\pgfplotsset{compat=1.18}
\usepgfplotslibrary{fillbetween}
\begin{document}
\begin{tikzpicture}
\begin{axis}[
xmin=0, xmax=25,
ymin=0, ymax=4,
ytick={1,2,3},
yticklabels={Condition 1, Condition 2, Condition 3},
]

\addplot[draw=none, name path=p1, y filter/.expression={1+0.1*y}] gnuplot[raw gnuplot] {plot "data1.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[draw=none, name path=m1, y filter/.expression={1-0.1*y}] gnuplot[raw gnuplot] {plot "data1.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[red!50] fill between [of=p1 and m1];

\addplot[draw=none, name path=p2, y filter/.expression={2+0.1*y}] gnuplot[raw gnuplot] {plot "data2.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[draw=none, name path=m2, y filter/.expression={2-0.1*y}] gnuplot[raw gnuplot] {plot "data2.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[green!50] fill between [of=p2 and m2];

\addplot[draw=none, name path=p3, y filter/.expression={3+0.1*y}] gnuplot[raw gnuplot] {plot "data3.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[draw=none, name path=m3, y filter/.expression={3-0.1*y}] gnuplot[raw gnuplot] {plot "data3.dat" using 1:2 smooth kdensity bandwidth 1.;};
\addplot[blue!50] fill between [of=p3 and m3];

\end{axis}
\end{tikzpicture}
\end{document}