If you want precise control of a document, LaTeX is a great way to go. It’s much more complicated than Markdown, but if you’re writing a journal article, thesis, or book, you may want the extra precision.

I’m assuming here that you already know how to write a LaTeX document. If you don’t, I’d suggesting focusing first on using knitr with markdown, study LaTeX elsewhere (I like this book, but it’s maybe not for beginners), and then coming back to this.

The first thing to note, with knitr and LaTeX, is that the file that mixes R code chunks and LaTeX should have the extension .Rnw. That stands for “R noweb”. Noweb was an early system (still used) for literate programming (mixing code and text).

Code chunk delimiters

Code chunks in .Rnw files are delimited with <<>>= at the top (you put the chunk label and any chunk options in the middle of that), and @ following. In-line code chunks are indicated with \Sexpr{} (you put the code inside the curly braces). Here’s a bit from my example:

We see that this is an intercross with \Sexpr{nind(sug)} individuals.
There are \Sexpr{nphe(sug)} phenotypes, and genotype data at
\Sexpr{totmar(sug)} markers across the \Sexpr{nchr(sug)} autosomes.  The genotype
data is quite complete.

Use {\tt plot()} to get a summary plot of the data.

<<summary-plot, fig.height=8>>=

Otherwise, everything about the code chunks and chunk options is the same as with R Markdown.


To produce LaTeX tables, you can use kable and xtable much as you would do with R Markdown. And the results look considerably nicer than the html tables. (You can’t use pander package within LaTeX.)

Here’s a kable example.

n <- 100
x <- rnorm(n)
y <- 2*x + rnorm(n)
out <- lm(y ~ x)
kable(summary(out)$coef, digits=2)

And here’s the xtable version. It’s simpler than the html version, where we assigned the output of xtable to an object and then printed it with type="html".

<<xtable, results="asis">>=
n <- 100
x <- rnorm(n)
y <- 2*x + rnorm(n)
out <- lm(y ~ x)
xtable(summary(out)$coef, digits=c(0, 2, 2, 1, 2))

Converting knitr/LaTeX to PDF


You can use RStudio to convert a .Rnw file to PDF and preview the result, in the same way you worked with R Markdown.

But the default in RStudio is still to use Sweave, so you first need to change that default. Go to the RStudio (on menu bar) → Preferences and select Sweave on the left. Then change the selection for “Weave Rnw files using:” from Sweave to knitr. You also have a choice of using pdfLaTeX or XeLaTeX.

Note: There is another place in RStudio that you can select knitr vs Sweave: Tools (on menu bar) → Project Options and select Sweave on left, then again change the selection for “Weave Rnw files using:” from Sweave to knitr. (From Tools on menu bar, you’ll also see “Global Options”; this seems to be the same as RStudio → Preferences.) It seems like the selection within Project Options overrides the selection within Global Options/Preferences.

Now, if you open a .Rnw file in RStudio, or if you create a new one with File → New → R Sweave, there will be a button “Compile PDF” that acts just like “Knit HTML” for R Markdown files. (Unfortunately, though, it’s not a cute button, like the yarn one.) Click that button, and a preview window will open.

Note: If you opened the new .Rnw file while the options were still set to Sweave (rather than knitr), the new file will have a line with \SweaveOpts{concordance=TRUE}. But if you’re using knitr, you’ll need to delete this line.


Yihui Xie recommends LyX, a graphical front-end for LaTeX that includes support for knitr. See Using knitr with LyX.


To convert a .Rnw file to pdf from the command-line (or within a GNU make file), you first use the knit function in the knitr package to process the code chunks and create a .tex file:

R -e 'library(knitr);knit("knitr_example.Rnw")'

You then the usual latex, pdflatex, or xelatex command to convert the .tex file to a PDF. If you need to submit sources to a journal, you can send them that intermediate .tex file that’s created.