3

I am new here so if there is anything Im not following the rules, please tell me. Im not a frequent user of pgfplots.

I am trying to connect coordinates in a boxplot with pgfplots and tikzpicture. The data is paired, every dot should connect to its counterpart as they are data of the same patient. Here you can see my result. enter image description here I do not understand why the dots are not connected. I found the coordinates of the points and put them in an array. At the end I connect de dots with draw lines. Does anyone know why some appear connected and others seem out of line? For anyone asking why I would use a boxplot, my supervisor explicatly asked for one, even though the data does ask for it. This picture gives a general idea of what I am trying to achieve within the boxplot:

enter image description here

\documentclass[a4paper,12pt,twoside]{book}
\usepackage{filecontents} 
\usepackage{graphicx, tabularx}
\usepackage{pgfplotstable}
\usepackage{pgfplots}
\pgfplotsset{compat=1.8}
\usepgfplotslibrary{statistics}
\usepackage{xcolor}
\begin{filecontents*}{data.csv}
b-SSFP_RA   b-SSFP_SA   CE_MRA_SA   CE_MRA_SA
246,78      288,75      254,99      307,63
283,38      183,85      283,56      60,35
158,01      205,85      170,87      88,10
114,81      322,72      107,00      28,42
161,04      169,25      129,56      69,58
\end{filecontents*}
\pgfplotstablegetrowsof{data.csv}
\pgfmathtruncatemacro{\N}{\pgfplotsretval-1}
\pgfplotsset{
  tick label style = {font=\sansmath\sffamily},
  every axis label = {font=\sansmath\sffamily},
  legend style = {font=\sansmath\sffamily},
  label style = {font=\sansmath\sffamily}}

\begin{document}

\begin{figure}
\centering
\begin{tikzpicture}
\begin{axis}[width = 0.9\textwidth,
boxplot/draw direction=y,
ylabel={CNR},
height=7cm,
very thick,
legend style={nodes=right,very thick},
boxplot={
draw position={1/5 + floor(\plotnumofactualtype/2) + 1/2*mod(\plotnumofactualtype,2)},
box extend=0.3},
ymajorgrids,
x=4cm,
xtick={0,1,2,...,10},
xticklabels={Renal artery,Segmental artery},%
x tick label as interval,
x tick label style={
text width=3.5cm,
align=center},
xlabel=Anatomic region,
legend entries = {b-SSFP, CE-MRA},
legend to name={legend},
name=border]


\addplot+ [blue,  opacity=1,fill=blue!30,fill opacity=0.3, very thick][row sep=\\,
     boxplot prepared={
     lower whisker=114.82,
     lower quartile=158.01,
     median=161.41,
     upper quartile=246.78,
     upper whisker=283.39,}
     ] coordinates {(0,247)(0,284)(0,158)(0,115)(0,161)}
     \foreach \i in {1,...,\N} {
                coordinate [pos=\i/\N] (a\i)
            };
\addplot+ [orange,  opacity=1,fill=orange!30,fill opacity=0.3, very thick][row sep=\\,
     boxplot prepared={
     lower whisker=169.26,
     lower quartile=183.85,
     median=205.85,
     upper quartile=288.75,
     upper whisker=322.72,}
     ] coordinates {(0,289)(0,184)(0,206)(0,323)(0,169)}
     \foreach \i in {1,...,\N} {
                coordinate [pos=\i/\N] (b\i)
            };
\addplot+ [blue,  opacity=1,fill=blue!30,fill opacity=0.3, very thick][row sep=\\,
    boxplot prepared={
     lower whisker=107.00,
     lower quartile=129.56,
     median=170.87,
     upper quartile=254.99,
     upper whisker=283.57,}
     ] coordinates {(0,255)(0,284)(0,171)(0,107)(0,130)}
          \foreach \i in {1,...,\N} {
                coordinate [pos=\i/\N] (c\i)
            };
\addplot+ [orange,  opacity=1,fill=orange!30,fill opacity=0.3, very thick][row sep=\\,
     boxplot prepared={
     lower whisker=28.42,
     lower quartile=60.35,
     median=69.58,
     upper quartile=88.10,
     upper whisker=307.63}
     ]  coordinates {(0,308)(0,60)(0,88)(0,28)(0,70)}
          \foreach \i in {1,...,\N} {
                coordinate [pos=\i/\N] (d\i)
            };
\end{axis}
 \foreach \i in {1,...,\N} {\draw (a\i) -- (b\i);
        };
 \foreach \i in {1,...,\N} {\draw (c\i) -- (d\i);
};
\node[below right] at (border.north east) {\ref{legend}};
\end{tikzpicture}
\end{figure}
\end{document}
4
  • Welcome to TEXSE, could you make your script compilable? Aug 7, 2019 at 9:12
  • Ive added packages etc.
    – MatthF
    Aug 7, 2019 at 9:18
  • @MatthF You are using an exotic class, probably many unneeded packages, and your document can't be compiled even after changing these because it relies on a file tables/CNR_Patienten.dat that you didn't provide! You can include it from your document using \RequirePackage{filecontents} \begin{filecontents*}{CNR_Patienten.dat} ... \end{filecontents*} for instance at the very beginning of your document (of course, first make sure it doesn't contain private data).
    – frougon
    Aug 7, 2019 at 10:01
  • Ive added the filecontents package and removed some most likely unneeded packages. Great thanks for the help. It should now run on its own.
    – MatthF
    Aug 7, 2019 at 10:11

1 Answer 1

2

In principle your idea to add the connecting lines is right, but does not work, because the outliers are drawn on an invisible line and thus the position is calculated on this line instead of the "index of the point". For details I refer to section 4.17.2 on page 356 of the PGFPlots manual (v1.16).

To circumvent this problem you can move the "outliers" to additional \addplot commands. For details have a look at the comments in the code.

% used PGFPlots v1.16
% changes to the data file
% - replaced commata by points
% - corrected/changed "SA" to "RA" in second last column (to avoid duplicate name)
\begin{filecontents*}{data.csv}
b-SSFP_RA   b-SSFP_SA   CE_MRA_RA   CE_MRA_SA
246.78      288.75      254.99      307.63
283.38      183.85      283.56      60.35
158.01      205.85      170.87      88.10
114.81      322.72      107.00      28.42
161.04      169.25      129.56      69.58
\end{filecontents*}
\documentclass[border=5pt]{standalone}
\usepackage{pgfplotstable}
    \usepgfplotslibrary{statistics}
    \pgfplotsset{
        compat=1.3,
        % create a style for the box plots
        % (which takes an argument for the color)
        box style/.style={
            #1,
            solid,
            fill=#1!30,
            fill opacity=0.3,
            boxplot={
                draw position={1/5 + floor(\plotnumofactualtype/2) + 1/2*mod(\plotnumofactualtype,2)},
                box extend=0.3,
            },
        },
        % create a style for the marks
        % (which also takes an argument for the color)
        mark style/.style={
            #1,
            mark=*,
            only marks,
            table/x expr={1/5 + floor(\plotnumofactualtype/2) + 1/2*mod(\plotnumofactualtype,2)},
        },
    }
    \pgfplotstablegetrowsof{data.csv}
    \pgfmathtruncatemacro{\N}{\pgfplotsretval-1}
\begin{document}
\begin{tikzpicture}
    \begin{axis}[
        width=0.9\textwidth,
        height=7cm,
        xlabel=Anatomic region,
        ylabel={CNR},
        xtick={0,...,3},
        xticklabels={Renal artery,Segmental artery},
        x tick label as interval,
        x tick label style={
            text width=3.5cm,
            align=center,
        },
        very thick,
        ymajorgrids,
        boxplot/draw direction=y,
        legend pos=outer north east,
        % because you (most likely) only want to plot the dots you need to
        % skip the boxplots, which can be done by giving an empty entry
        % (this is why the entries start with a comma)
        legend entries={
            ,b-SSFP,
            ,CE-MRA
        },
    ]
        \addplot+ [
            % use the created style here
            box style=blue,
            boxplot prepared={
                lower whisker=114.82,
                lower quartile=158.01,
                median=161.41,
                upper quartile=246.78,
                upper whisker=283.39,
            },
        % remove the marks from here ...
        ] coordinates {};
        % ... and add them as extra `\addplot`s ...
        \addplot [
            mark style=blue,
        ] table [
            y=b-SSFP_RA,
        ] {data.csv}
            % ... including the `\foreach` part giving coordinate names to
            % the points
            % (Please note that the index starts at 0 and not 1.)
            \foreach \i in {0,...,\N} {
                coordinate [pos=\i/\N] (a\i)
            }
        ;

        \addplot+ [
            box style=orange,
            boxplot prepared={
                lower whisker=169.26,
                lower quartile=183.85,
                median=205.85,
                upper quartile=288.75,
                upper whisker=322.72,
            },
        ] coordinates {};
        \addplot [
            mark style=orange,
        ] table [
            y=b-SSFP_SA,
        ] {data.csv}
            \foreach \i in {0,...,\N} {
                coordinate [pos=\i/\N] (b\i)
            }
        ;

        \addplot+ [
            box style=blue,
            boxplot prepared={
                lower whisker=107.00,
                lower quartile=129.56,
                median=170.87,
                upper quartile=254.99,
                upper whisker=283.57,
            },
        ] coordinates {};
        \addplot [mark style=blue] table [
            y=CE_MRA_RA,
        ] {data.csv}
            \foreach \i in {0,...,\N} {
                coordinate [pos=\i/\N] (c\i)
        };

        \addplot+ [
            box style=orange,
            boxplot prepared={
                lower whisker=28.42,
                lower quartile=60.35,
                median=69.58,
                upper quartile=88.10,
                upper whisker=307.63,
            },
        ] coordinates {};
        \addplot [mark style=orange] table [
            y=CE_MRA_SA,
        ] {data.csv}
            \foreach \i in {0,...,\N} {
                coordinate [pos=\i/\N] (d\i)
            }
        ;
    \end{axis}

    % draw the connecting lines
    \foreach \i in {0,...,\N} {
        \draw (a\i) -- (b\i);
        \draw (c\i) -- (d\i);
    }
\end{tikzpicture}
\end{document}

image showing the result of above code

8
  • I like your code with your result. It is into my favorities.
    – Sebastiano
    Aug 7, 2019 at 12:17
  • 1
    Thanks for the flowers :) Aug 7, 2019 at 12:33
  • Thank you! This works wonderfully!
    – MatthF
    Aug 7, 2019 at 13:55
  • For some reason I cannot get the old legend to work, with the node at the end. Do you have any tips to get the legend to work again? The \ref is unrecognized for some reason.
    – MatthF
    Aug 7, 2019 at 15:32
  • Edit from previous For some reason I cannot get the old legend to work, with the node at the end. Do you have any tips to get the legend to work again? I think it has something to do with the node position. I got to be honest, this plot is getting above my head tbh
    – MatthF
    Aug 7, 2019 at 15:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .