8

I am trying to plot a scatterplot with a fit line that I have already calculated, but I am having trouble getting them to line up and use the same scale. I tried doing it with

\begin{tikzpicture}[domain=0:8]
\begin{axis}[xlabel={$\Delta x$ (m)}, ylabel={$Mg$ (N)}]
    \draw[color=red, domain=0:1000] plot (\x,{0.0051026+10.372231*\x}) node[below right] {};
\addplot[scatter, only marks, scatter src=\thisrow{class},
      error bars/.cd, y dir=both, x dir=both, y explicit, x explicit, error bar style={color=mapped color}]
      table[x=x,y=y,x error=xerr,y error=yerr] {
x      xerr      y        yerr     class
0.0047 0.0007071 0.054039 0.000098 0
0.0142 0.0007071 0.152651 0.000098 0
0.0237 0.0007071 0.252051 0.000098 0
0.0332 0.0007071 0.350466 0.000098 0
0.0525 0.0007071 0.548380 0.000098 0
0.0622 0.0007071 0.646893 0.000098 0
0.0720 0.0007071 0.746195 0.000098 0
0.0802 0.0007071 0.844709 0.000098 0
};
\end{axis}
\end{tikzpicture}

but I am getting a plot that looks like

enter image description here

Can I make the fit line match up with the scatter plot with the type of plot I am using? If not, how can I make this work?

murgatroid99
  • 515
  • Maybe useful: http://tex.stackexchange.com/q/11251/14100 – Scott H. Oct 10 '12 at 03:15
  • That does look useful, but I'd prefer to be able to input a function that I already have – murgatroid99 Oct 10 '12 at 03:31
  • As per the question @ScottH. linked to, adding a separate \addplot [blue, mark=none] table[y={create col/linear regression={y=y}}] { with the same data will create a linear regression line to fit the given data. If you want to use the function you already computed you need to use a separate \addplot. – Peter Grill Oct 10 '12 at 04:03

1 Answers1

10

You have two issues with your MWE:

  • You are trying to mix tikz macros within pgfplot's axis environment.
  • Your domain for the function you are tying to plot is incorrect.

1. Manual Fit Line:

So using another \addplot and correcting the domain to 0:08:

\addplot [color=red, domain=0:0.08, mark=none] {0.0051026+10.372231*\x};

you get:

enter image description here

Code:

\documentclass{article}
\usepackage{pgfplots}
\usepackage{pgfplotstable}

\begin{document}
\begin{tikzpicture}[domain=0:8]
\begin{axis}[xlabel={$\Delta x$ (m)}, ylabel={$Mg$ (N)}]
%    \draw[color=red, domain=0:1000] plot (\x,{0.0051026+10.372231*\x}) node[below right] {};


\addplot [color=red, domain=0:0.08, mark=none] {0.0051026+10.372231*\x};


\addplot[scatter, only marks, scatter src=\thisrow{class},
      error bars/.cd, y dir=both, x dir=both, y explicit, x explicit, error bar style={color=mapped color}]
      table[x=x,y=y,x error=xerr,y error=yerr] {
x      xerr      y        yerr     class
0.0047 0.0007071 0.054039 0.000098 0
0.0142 0.0007071 0.152651 0.000098 0
0.0237 0.0007071 0.252051 0.000098 0
0.0332 0.0007071 0.350466 0.000098 0
0.0525 0.0007071 0.548380 0.000098 0
0.0622 0.0007071 0.646893 0.000098 0
0.0720 0.0007071 0.746195 0.000098 0
0.0802 0.0007071 0.844709 0.000098 0
};


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

2. Automatic Fit Line:

Alternatively you could let pgfplots compute the regression line for you by using the same data with:

\addplot [red, mark=none] table[y={create col/linear regression={y=y}}

enter image description here

Notes:

  • Note that in this case the regression line is automatically only between the extreme points. You would need to adjust your domain to get the same results.
  • You should not duplicate the data as I have done.

Code:

\documentclass{article}
\usepackage{pgfplots}
\usepackage{pgfplotstable}

\begin{document}
\begin{tikzpicture}[domain=0:8]
\begin{axis}[xlabel={$\Delta x$ (m)}, ylabel={$Mg$ (N)}]
%    \draw[color=red, domain=0:1000] plot (\x,{0.0051026+10.372231*\x}) node[below right] {};


\addplot[scatter, only marks, scatter src=\thisrow{class},
      error bars/.cd, y dir=both, x dir=both, y explicit, x explicit, error bar style={color=mapped color}]
      table[x=x,y=y,x error=xerr,y error=yerr] {
x      xerr      y        yerr     class
0.0047 0.0007071 0.054039 0.000098 0
0.0142 0.0007071 0.152651 0.000098 0
0.0237 0.0007071 0.252051 0.000098 0
0.0332 0.0007071 0.350466 0.000098 0
0.0525 0.0007071 0.548380 0.000098 0
0.0622 0.0007071 0.646893 0.000098 0
0.0720 0.0007071 0.746195 0.000098 0
0.0802 0.0007071 0.844709 0.000098 0
};


\addplot [red, mark=none] table[y={create col/linear regression={y=y}}] {
x      xerr      y        yerr     class
0.0047 0.0007071 0.054039 0.000098 0
0.0142 0.0007071 0.152651 0.000098 0
0.0237 0.0007071 0.252051 0.000098 0
0.0332 0.0007071 0.350466 0.000098 0
0.0525 0.0007071 0.548380 0.000098 0
0.0622 0.0007071 0.646893 0.000098 0
0.0720 0.0007071 0.746195 0.000098 0
0.0802 0.0007071 0.844709 0.000098 0
};


\end{axis}
\end{tikzpicture}
\end{document}
Community
  • 1
Peter Grill
  • 223,288
  • What is the class column for? – Magpie Jan 28 '13 at 23:21
  • @Magpie: Don't know. That was directly from the one provided in the MWE so assume that it is intended to be another column of data. – Peter Grill Jan 29 '13 at 00:40
  • Is it possible to get the automatic regression line labelled on the the graph too automatically? And include an R^2 value? – HCAI Mar 13 '13 at 16:45
  • @user1134241: This might be better as a separate question. Also, what do you mean by "include a R^2 value?" – Peter Grill Mar 13 '13 at 18:07
  • @PeterGrill Ok I'll open a question up. The $R^2$ value is the correlation coefficient, ie the square of the average distance from the data to the line of best fit. I noticed though that you said that the fit is made through with the extreme values and not actually a linear regression, or have I misunderstood? – HCAI Mar 13 '13 at 19:09
  • 1
    @user1134241: The line is drawn only between the extreme values (as it should be) and not extended past them, the computation is AFAIK a complete linear regression. – Peter Grill Mar 13 '13 at 19:14