将线性回归的斜率附加到 pgfplotstable 列

Question

假设我有一个带有多个数据系列的 pgfplots 轴。现在我循环遍历它并计算每个系列的线性回归。如何将回归线的斜率存储到 pgfplots 表列中？这是一个（非工作）最小示例：

\documentclass{article}

\usepackage{pgfplots,pgfplotstable}

\pgfplotsset{
  compat=1.13
  }

\begin{document}

%% Dummy data
\pgfplotstableread{
  x1  y1 x2  y2 
  1   1  1   2 
  2   2  2   4 
  3   3  3   6 
  4   4  4   8 
}{\mytable}


  \begin{tikzpicture}
    \begin{axis}[width=\textwidth,height=8cm]
      \pgfplotsinvokeforeach{1,2}{
        \addplot+[only marks,] table[x=x#1,y=y#1]  {\mytable};
        \addplot+[smooth,no markers] table[x=x#1,y={create col/linear regression={y = y#1}}]
        {\mytable};
        \xdef\slope##1{\pgfplotstableregressiona}
    }
  \end{axis}     
  \end{tikzpicture}

%% calling \slope1 and \slope2 doesn't seem to be a good way to do this, since it should automatically work independently of how many slopes I calculate. However, it doesn't work anyway.    
\pgfplotstableread{
  Nr  slope
  1 \slope1
  2 \slope2
}{\atable}  

\pgfplotstableset{
columns/slope/.style={column name=Slope},
}



\pgfplotstabletypeset[
columns={Nr,slope},
columns/Nr/.style={column type=l,column name=Nr},
]{\atable}



\end{document}

输出：

顺便说一句，我不知道这个函数的##1作用是什么。我只是偶然用了它，然后发现它可以编译（但它不会在表格中打印正确的斜率）。

Answer 1

一种选择是使用 来etoolbox存储\csxdef值并\csuse检索它：

＃＃代码：

\documentclass{article}

\usepackage{pgfplots,pgfplotstable}
\usepackage{etoolbox}

\pgfplotsset{
  compat=1.13
  }

\begin{document}

%% Dummy data
\pgfplotstableread{
  x1  y1 x2  y2 
  1   1  1   2 
  2   2  2   4 
  3   3  3   6 
  4   4  4   8 
}{\mytable}


  \begin{tikzpicture}
    \begin{axis}[width=\textwidth,height=8cm]
      \pgfplotsinvokeforeach{1,2}{
        \addplot+[only marks,] table[x=x#1,y=y#1]  {\mytable};
        \addplot+[smooth,no markers] table[x=x#1,y={create col/linear regression={y = y#1}}]
        {\mytable};
        \csxdef{slope #1}{\pgfplotstableregressiona}
    }
  \end{axis}     
  \end{tikzpicture}

%% calling \slope1 and \slope2 doesn't seem to be a good way to do this, since it should automatically work independently of how many slopes I calculate. However, it doesn't work anyway.    
\pgfplotstableread{
  Nr  slope
  1 \csuse{slope 1}
  2 \csuse{slope 2}
}{\atable}  

\pgfplotstableset{
columns/slope/.style={column name=Slope},
}



\pgfplotstabletypeset[
columns={Nr,slope},
columns/Nr/.style={column type=l,column name=Nr},
]{\atable}

\end{document}