[Tex/LaTex] siunitx: What is the correct way to output mean and standard deviation as “M (SD)” in an S-column

rsiunitx

I am trying to use siunitx in a psychology results table to line up decimals, and the table looks like this:

\documentclass[]{article}
\begin{document}
\begin{table}[h]
    \begin{tabular}{@{}cc@{}}
    \multicolumn{2}{c}{Mean (SD)} \\
    \hline
    A & B \\
    \hline
    .45 (.23) & .78 (.11) \\ 
    \end{tabular}
\end{table}
\end{document}

This does not follow the short form uncertainty that siunitx uses, rather it gives the full standard deviation value (as is usual in psychology).

My question is how can I make siunitx output to the format M (SD) whilst providing SD with an actual standard deviation value in the code (such as .11)? I cannot use the short uncertainty format of siunitx because the input is from R code (using knitr) that returns the real standard deviation value, e.g. \Sexpr{sd(some.numbers)}.

Edit: I have tried separate-uncertainty but this seems to force an output of M ± SD and I would like to use M (SD).

Best Answer

The formats parsed by siunitx are 1.23(4) and 1.23 \pm 0.04, and no others. Thus to achieve the desired effect some parsing has to be done separately. One approach is to grab the cell contents, alter the formatting and use the \tablenum command to do the alignment

\documentclass[]{article}
\usepackage{siunitx}
\def\converter\ignorespaces#1(#2){%
  \begingroup\tablenum[table-format = 1.2]{#1}\endgroup\space
  (\tablenum[table-format = 1.2]{#2})}
\newcolumntype{R}{>{\converter}c}
\begin{document}
\begin{table}[h]
    \begin{tabular}{@{}RR@{}}
    \multicolumn{2}{@{}c@{}}{Mean/SD} \\
    .45 (.23) & .78 (.11) \\
    0.2(.4) \\ 
    \end{tabular}
\end{table}
\end{document}

Related Solutions

[Tex/LaTex] How to denote the standard deviation with siunitx

The code you are looking for is on page 43 of the siunitx manual (as of 2011/10/14):

When a number has multiple parts (such as a separate uncertainty) then the unit must apply to all parts of the number. How this is shown is controlled using the multi-part-units options. The standard setting is brackets, which will place the entire numerical part in brackets and use a single unit symbol. Alternative options are repeat (print the unit for each part of the number) and single (print only one unit symbol: mathematically incorrect).

\sisetup{separate-uncertainty}%
\SI{12.3(4)}{\kilo\gram} \\
\SI[multi-part-units = brackets]{12.3(4)}{\kilo\gram} \\
\SI[multi-part-units = repeat]{12.3(4)}{\kilo\gram} \\
\SI[multi-part-units = single]{12.3(4)}{\kilo\gram}

[Tex/LaTex] Building TeX file from knitr Rnw as child of larger parent TeX file to use R objects and output

knitr has a few pretty straightforward ways of handling this.

Option 1: Using `knit_child()` with inline R code

Say your setup is like the following. In the same directory, you have:

`graph.R`

## ---- graph
library(ggplot2)
CarPlot <- ggplot() +
    stat_summary(data= mtcars,
                 aes(x = factor(gear),
                     y = mpg
                 ),
                 fun.y = "mean",
                 geom = "bar"
                 )
CarPlot

`chapter1.Rnw`

Hey, look, a graph (Figure~\ref{fig:graph})!

<<graph, echo=FALSE, message=FALSE, fig.lp='fig:', out.width='.5\\linewidth', fig.align='center', fig.cap="A graph", fig.pos='h!'>>=
@

`main.Rnw`

\documentclass{article}

\begin{document}

<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@

\Sexpr{knit_child('chapter1.Rnw')}

\end{document}

Then, you can knit the main.Rnw file and compile the resulting .tex file with either pdflatex or xelatex.

The output is:

enter image description here

Note that you can also read the external .R file from the child .Rnw file.

So, the following would have worked just as well.

`chapter1-mod.Rnw`

<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@

Hey, look, a graph (Figure~\ref{fig:graph})!

<<graph, echo=FALSE, message=FALSE, fig.lp='fig:', out.width='.5\\linewidth', fig.align='center', fig.cap="A graph", fig.pos='h!'>>=
@

`main-mod.Rnw`

\documentclass{article}

\begin{document}

\Sexpr{knit_child('chapter1-mod.Rnw')}

\end{document}

Option 2: Using chunk option `child`

Assuming you have graph.R and chapter1.Rnw from above in the same directory, then your main.Rnw should be:

\documentclass{article}

\begin{document}

<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@

<<child-demo, child='chapter1.Rnw'>>=
@

\end{document}

Note that you can also read the external .R file from within the child document in this case, too.

So, assuming you had graph.R and chapter1-mod.Rnw from above in the same directory, then your main-mod.Rnw file should be:

\documentclass{article}

\begin{document}

<<child-demo, child='chapter1-mod.Rnw'>>=
@

\end{document}

Best Answer

Related Solutions

[Tex/LaTex] How to denote the standard deviation with siunitx

[Tex/LaTex] Building TeX file from knitr Rnw as child of larger parent TeX file to use R objects and output

Option 1: Using knit_child() with inline R code

graph.R

chapter1.Rnw

main.Rnw

chapter1-mod.Rnw

main-mod.Rnw

Option 2: Using chunk option child

Related Question