I have a table of estimated averages and their associated confidence intervals. But when I plot the confidence intervals, pgfplots
treats the interval coordinates as error values. Is there a way to tell pgfplots
to plot the CI data as is, instead of adding/subtracting the CI values from the mean?
MWE
\documentclass[border=2pt]{standalone}
\usepackage{pgfplots}
\usepackage{filecontents}
\begin{filecontents*}{data.txt}
Round NUM CI-L CI-H
1 1.50 1.09 1.91
2 1.05 0.61 1.49
3 0.50 0.14 0.86
4 0.20 -0.03 0.43
5 0.35 -0.15 0.85
6 0.30 -0.05 0.65
7 0.20 -0.10 0.50
8 0.25 -0.15 0.65
9 0.20 -0.10 0.50
10 0.20 -0.10 0.50
11 0.15 -0.06 0.36
12 0.20 -0.10 0.50
13 0.20 -0.10 0.50
14 0.20 -0.10 0.50
15 0.20 -0.10 0.50
16 0.25 -0.15 0.65
17 0.15 -0.06 0.36
18 0.20 -0.10 0.50
19 0.20 -0.10 0.50
20 0.15 -0.06 0.36
\end{filecontents*}
\begin{document}
\begin{tikzpicture}
\begin{axis}
\addplot table [x index=0, y index=1] {data.txt};
\addplot[error bars/.cd,y dir=minus,y explicit] table [x index=0, y index=1,y error index=2] {data.txt};
\addplot[error bars/.cd,y dir=plus,y explicit] table [x index=0, y index=1,y error index=3] {data.txt};
\end{axis}
\end{tikzpicture}
\end{document}
Best Answer
This can be done using
y error plus expr
andy error minus expr
keys. Also you can combine all the three\addplot
commands to one.