\documentclass{article}
\usepackage{booktabs}
\usepackage{longtable}
\usepackage{capt-of}

\setlength{\parskip}{1em}
\setlength{\parindent}{0em}

%\VignetteEngine{knitr::knitr}
%\VignetteIndexEntry{Using papeR with LaTeX}

<<include = FALSE>>=
library("knitr")
opts_chunk$set(message = FALSE, warning = FALSE, echo = TRUE, results = 'asis')
data(Orthodont, package = "nlme")
@

\title{Using papeR with \LaTeX}
\author{Benjamin Hofner}
\date{Version \Sexpr{packageDescription("papeR")$Version}} %$

\begin{document}
\maketitle

\begin{abstract}
  \setlength{\parskip}{0.5em}

  The main goal of the package \textbf{papeR} is to ease statistical reporting
  and thus to ease reproducible research. By relying on powerful tools such as
  the \texttt{Sweave} command kit, or the packages \textbf{knitr} and
  \textbf{xtable}, the package can be easily integrated in existing workflows.

  The package provides an infrastructure to handle variable labels which are
  used in all other functions (\texttt{labels()}), allows to create (complex)
  summary tables of the data sets (\texttt{summarize()}) and to easily plot the
  data (\texttt{plot()} for labeled \texttt{data.frame}s), and enhances summary
  tables of statistical models by (possibly) adding confidence intervals,
  significance stars, odds ratios, etc. and by separating variable names and
  factor levels (\texttt{prettify()}).
\end{abstract}

\section{Introduction}

This is a short description of some of the \LaTeX related features of
\texttt{papeR}. For installation instructions and a comprehensive overview of
the features of \texttt{papeR} see also the vignette on using \texttt{papeR} to
(mainly) produce Markdown output (e.g. via
\texttt{vignette("papeR\_introduction", package = "papeR")}).

In short, we load the package, load an example data set, and set variable labels:
<<>>=
library("papeR")
data(Orthodont, package = "nlme")
labels(Orthodont) <- c("fissure distance (mm)",
                       "age (years)", "Subject", "Sex")
@

\section{Pretty tables}

To produce \LaTeX tables, we heavily rely on the R package \textbf{xtable}. Note
that all arguments to \texttt{xtable()} or \texttt{print.xtable()} can be used
as usual, yet, some defaults were changed in package \textbf{papeR}. For
example, we use per default the \LaTeX package \texttt{booktabs} to print
tables and tables are not set as floating objects. For further changes see the
manual of \texttt{xtable.summary}.

\subsection{Summary tables for numerical variables}

Now we can produce summary tables for numeric variables:
<<>>=
xtable(summarize(Orthodont, type = "numeric"))
@

Grouped statistics with tests can be obtained via:
<<>>=
xtable(summarize(Orthodont, type = "numeric", group = "Sex"))
@

Per default, t-tests are computed. To change the test, one can use:
<<>>=
xtable(summarize(Orthodont, type = "numeric", group = "Sex",
                 test = c("wilcox.test", "t.test")))
@
To use Wicoxon tests for all variables, one could simply set \texttt{test =
  "wilcox.test"}, or one could switch off tests by setting \texttt{test =
  FALSE}.

To drop some of the statistics one can set several options to \texttt{FALSE}.
E.g., if we do not want to show the five-number summaries (minimum, 25\%
quantile, median, 75\% quantile, maximum), one can use
<<>>=
xtable(summarize(Orthodont, type = "numeric", group = "Sex",
                 quantiles = FALSE))
@
Alternatively or additionally, one could also drop $N$ (\texttt{count = FALSE})
or mean and standard deviation (\texttt{mean\_sd = FALSE}). For details see also
the manual of \texttt{summary\_numeric()}.

\subsection{Summary tables for factor variables}

In the same way, summary tables for factors can be computed. Here, we only want
to print the variable \texttt{Sex}:
<<>>=
xtable(summarize(Orthodont, type = "factor", variables = "Sex"))
@

If tables are longer than one page or cross a page, one can also use
\texttt{"tabular.environment = "longtable""}:
<<>>=
print(xtable(summarize(Orthodont, type = "factor")),
      tabular.environment = "longtable")
@
which automatically specifies the table header such that it is repeated at the
top of each new page.

To additionally obtain the cumulative frequencies, we can use:
<<>>=
xtable(summarize(Orthodont, type = "factor", variables = "Sex",
                 cumulative = TRUE))
@

As for numerical summaries, grouped statistics are tested. Per default, Fisher's
exact test is used. To speed up computations (of this non-sense test), we only
use a small subset of the original data:
<<>>=
Ortho_small <- subset(Orthodont,
                      Subject %in% c("M01", "M02", "F01", "F02"))
xtable(summarize(Ortho_small, type = "factor",
                 variables = "Subject", group = "Sex"))
@

\subsection{Captions}

As usual, all floating tables can have captions. Per default, these are printed
above the table. Note that by using the \LaTeX package \texttt{capt-of}, one can
specify table captions and labels also for non floating tables as shown in
Table~\ref{tab:Fisher}.
<<>>=
xtable(summarize(Ortho_small, type = "factor",
                 variables = "Subject", group = "Sex"),
       caption = "Example table for Fisher's exact test",
       label = "tab:Fisher")
@

\end{document}