supplement_with_talk2stat.tex

\documentclass[12pt]{amsart}
\usepackage[top=1.5in,bottom=1.5in,left=1in,right=1in]{geometry}
\usepackage[R]{runcode}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\title{Minimum working examples for\\ Reproducible Science with \LaTeX\space (`server mode')}


\begin{document}

\maketitle

\begin{center}
\author{Haim Bar and HaiYing Wang\\ University of Connecticut}\\
\today
\end{center}

This supplement provides minimum working examples in R for the procedure proposed in the main paper using the ``server mode'' provided by the Python \textit{talk2stat} package. Please make sure you have Python3 available on your computer. Install the \textit{talk2stat} package by the following command line.
\begin{verbatim}
  pip3 install talk2stat
\end{verbatim}
If the Python package \textit{pygments} (for grammar highlighting) is not available on your computer please install it,  or declare the `nominted' option, namely, use
\begin{verbatim}
  \usepackage[nominted,R]{runcode}
\end{verbatim}
to load the \textit{runcode} package. This will use the \textit{fvextra} package which does not provide syntax highlights. If \textit{pygments} is available, use
\begin{verbatim}
  \usepackage[R]{runcode}
\end{verbatim}
to load the \textit{runcode} package. The `R' option will start an R server.\\

The program Code/code1.R generates a vector \texttt{x} by drawing a random sample of size 100 from a standard normal distribution, and generates \texttt{y} as \texttt{1+x+$\epsilon$}, where the error term \texttt{$\epsilon$}
is also drawn from a standard normal distribution.
Then, we fit a linear model, \texttt{y$\sim$x}. To show the source file's contents, we include the following in the tex document:
\begin{verbatim}
  \showCode{R}{Code/code1.R}
\end{verbatim}
which produces the following:
\showCode{R}{Code/code1.R}
To show only lines 2-5 of the source code, we use
\begin{verbatim}
  \showCode{R}{Code/code1.R}[2][5]
\end{verbatim}
which produces the following:
\showCode{R}{Code/code1.R}[2][5]
\vspace{2mm}
% \section{Run source code and include output}

To execute the source code in Code/code1.R, we put the following in the tex file:
\begin{verbatim}
  \runR{Code/code1.R}{fitLinear}
\end{verbatim}
\runR{Code/code1.R}{fitLinear}
We include the output by using
\begin{verbatim}
  \includeOutput{fitLinear}
\end{verbatim}
and we obtain the following result:
\includeOutput{fitLinear}

In the above example, if the second argument in \verb|\runR| is empty, then the second argument in \verb|\includeOutput| should also be empty, namely, use
\begin{verbatim}
  \runR{Code/code1.R}{}
  \includeOutput{}
\end{verbatim}
to include the above output. We recommend avoiding this type of usage because \verb|\includeOutput| shows the results from the latest execution of \verb|\runR| with empty third argument, and this may make referencing harder to manage.\\

With the server mode, the R server is continuously running and all variables are always available, so we can use new codes for continuous calculations. For example, we want to use the variable \verb|fit| to create an ANOVA table and calculate the mean squared error (MSE) using the code in Code/code2.R:
\showCode{R}{Code/code2.R}
We use
\begin{verbatim}
  \runR{Code/code2.R}{linearANOVA}
  \includeOutput{linearANOVA}[tex]
\end{verbatim}
to obtain:
\runR{Code/code2.R}{linearANOVA}
\includeOutput{linearANOVA}[tex]\\
Here the code produces pure latex output, so we use the assign the value `tex' to the optional second argument of \verb|\includeOutput| to include the output as-is. \\

\begin{filecontents*}{tmp/linearMSE.R}
cat(format(sum(fit$residuals^2)/fit$df.residual, digit=2))
\end{filecontents*}
\runExtCode{Rscript --save --restore}{tmp/linearMSE.R}{linearMSE}{}
To embed R code in tex file and include the output in line, we use the \verb|\inlnR| command. For example, we use the following in the tex file to obtain the result: ``The MSE is \inlnR{```cat(MSE)```}''.
\begin{verbatim}
  The MSE is \inlnR{```cat(MSE)```}
\end{verbatim}
The \verb|\inlnR| command is also useful to show other types of output from previous calculations. As an example, we use 
\begin{verbatim}
  \inlnR{```aov(fit)```}[vbox]
\end{verbatim}
to obtain the following result.
\inlnR{```aov(fit)```}[vbox]

We can embed more complicated R code in tex file using \LaTeX\space \textit{filecontents} environment. For instance, the following code in a tex file will create an R script file plot.R in the tmp folder, and then implement it to create a pdf figure tmp/linearScatter.pdf.
\begin{verbatim}
 \begin{filecontents*}{tmp/plot.R}
  pdf("tmp/linearScatter.pdf", width=6, height=4)
  plot(x, y, pch=19, col="red" ,cex=0.8)
  dev.off()
 \end{filecontents*}
 \runR{tmp/plot.R}{linearScatter}
\end{verbatim}
\begin{filecontents*}{tmp/plot.R}
  pdf("tmp/linearScatter.pdf", width=6, height=4)
  plot(x, y, pch=19, col="red" ,cex=0.8)
  dev.off()
\end{filecontents*}
\runR{tmp/plot.R}{linearScatter}
We can include the resulting figure using \verb|\includegraphics{}| and the \textit{figure} environment in the usual way, e.g., with
\begin{verbatim}
  \begin{figure}
  \centering
  \includegraphics[scale=0.7]{tmp/linearScatter.pdf}
  \caption{A scatter plot of the simulated data}
  \end{figure}
\end{verbatim}
This produces Figure~\ref{fig:1}.

\begin{figure}[h]
\centering
\includegraphics[scale=0.7]{tmp/linearScatter.pdf}
\caption{A scatter plot of the simulated data}\label{fig:1}
\end{figure}

Note that the working directory of the R server section is the same as that of the main tex file, so in the above example we can use relative directory for the pdf figure (\verb|"tmp/linearScatter.pdf"|). If the file locations are changed after the R server section starts, one needs to use an absolute directory for the pdf figure, or one needs to end the R server section using the following command.
\begin{verbatim}
python3 -c 'from talk2stat.talk2stat import client; client("./","R","QUIT")'
\end{verbatim}
After that, a compilation of the tex file in the new location will stat an R server section with the working directory as the new location.


\bibliographystyle{unsrt}
\bibliography{refs}

\end{document}