Many of you are using Sweave for making statistical reports. I thought it might
be helpful to some to see an example of the setup I use (in Linux). For those
of you who have not yet discovered the power and productivity gains from using
Fritz Leisch's wonderful package, I encourage you to give Sweave a try.
%File: model.nw
%Usage:
% Put library(Hmisc;Design;tools} in .First
% Sweave model - runs model.nw to produce model.tex and
% graphics files in graphics/
% Sweave shell script defined as echo "Sweave(\"$1.nw\")" |
R --no-save
% latex or pdflatex model; bibtex model; latex or pdflatex model
% To get .R file: Stangle model
\documentclass{article}
\usepackage{relsize,setspace} % used by latex(describe( ))
\newcommand{\co}[1]{\texttt{\smaller #1}}
\title{Analysis}
\author{Frank E Harrell Jr}
\begin{document}
\SweaveOpts{prefix.string=graphics/plot}
\setkeys{Gin}{width=1.0\textwidth}
\maketitle
<<echo=F>>Hmisc.version <-
package.description('Hmisc')['Version']
Design.version <- package.description('Design')['Version']
@
All calculations were done using \Sexpr{R.version.string} \cite{Roriginal} on
RedHat Linux 9.0, using version \Sexpr{Hmisc.version} of the \texttt{Hmisc}
package and version \Sexpr{Design.version} of the \texttt{Design} package.
\section{Descriptive Statistics}
<<results=tex>>load('mydata.sav')
latex(describe(mydata), file='') # generate LaTeX code in place
@
The variable clustering diagram below shows which variables in the
dataset are interrelated. The $y$-axis is the Spearman $\rho^2$
(squared rank correlation coefficient). \co{addon} and
\co{mica} cannot be considered together because of an undefined
correlation between them. \co{addon} correlates weakly with
\co{sodm} (only). \co{usedlabd} is always unity when it is not
missing, so it is also omitted from variable clustering.
\begin{center}
<<fig=T>>plot(varclus(~.-addon-usedlabd, data=mydata))
@
\end{center}
\section{Analysis of Dose Effects}
. . . .
\bibliography{/home/feh/bib/feh.bib}
\bibliographystyle{abbrv}
\end{document}
For more information about reproducible statistical reporting, literate
programming, and Sweave see
http://hesweb1.med.virginia.edu/biostat/s/LiveDoc.html
---
Frank E Harrell Jr Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine http://hesweb1.med.virginia.edu/biostat
