From CommunityData
Jump to: navigation, search

Knitr is a package that allows you to import dataframes from R and dynamically generate LaTeX tables in your paper. Using knitr is good practice for times when you're working on a paper and make a slight change to your analysis, and need to update every instance where you've reported results.


First, you'd want to create an R script that's organized around every table you want to create in your document. The script should create separate dataframes for every one of them. If your table is going to have rownames, include a vector of rownames in your data frame.

A data frame for a summary stats table could look something like this, for instance (assuming you've loaded your data in a dataframe called d):

summary.stats <- data.frame(cbind(
                             c("Var 1", "Var 2", "Var 3"),
                             c(min(d$var.1), min(d$var.2), min(d$var.3)),
                             c(median(d$var.1), median(d$var.2), median(d$var.3)),
                             c(max(d$var.1), max(d$var.2), max(d$quality.score))

Make sure to also include column names for your dataframe:

names(summary.stats) <- c("Measures", "Min.", "Median", "Max.")

Finally, append each table dataframe to a list (i.e. results$summary.stats <- summary.stats), and save this list as an .Rdata file. You'd probably also want to store your entire dataset in the results list (i.e. as results$d ).

Latex Preamble[edit]

Save the .Rdata file in the same folder as your tex document. Now, include a preamble like this in your tex file (by all your \usepackage commands):

<<init, echo=FALSE, message=FALSE>>=
knit_hooks$set(document = function(x) {
'\\usepackage[usenames,dvipsnames]{color}', x, fixed = TRUE)

Then, after your \begin{document} command, import your dataset and necessary R packages. You'll need xtable for generating LaTeX tables from Rdata files and ggplot2 for charts and diagrams.

 <<import, echo=FALSE, message=FALSE>>=
 bold <- function(x) {paste('{\\textbf{',x,'}}', sep =)}
 gray <- function(x) {paste('{\\textcolor{gray}{',x,'}}', sep =)}
 wrapify <- function (x) {paste("{", x, "}", sep="")}
 f <- function (x) {formatC(x, format="d", big.mark=',')}

Using xtable and inserting S expressions[edit]


Each time you want to generate LaTeX code for a table in your document, you need to use knitr to create an R environment and use the xtable package. For instance, printing the summary stats table from the example above would look like this:

<<summary stats, echo=FALSE, message=FALSE, results='asis'>>=
     caption =
     "Summary statistics.",
     label="table:descriptives", align="llrrr"),

Note that there's a funny thing going on with the align parameter here. R includes some default rownames for dataframes (usually something like X1, X2 etc). Since we have a column that contains our rownames, we need to suppress the default rownames by specifying include.rownames=FALSE. However, even though the summary.stats dataframe contained only four columns, we still need to specify five column alignments in for the align parameter because the first one corresponds to the default rownames column in the R dataframe. Not specifying the extra column alignment parameter will give you an error.

Inserting S expressions[edit]

You also will want to have all the instances where you mention things like estimates or quantities from your dataset within the text body to be automatically generated every time you compile your TeX file. In order to do that, every time you mention a number in the text, you need to replace it with an \Sexpr{} command.

For instance, if your text contains a line that says something like "The median of Var 1 is x", then you replace x with:


The "f" function here is a formatting function defined in the preamble (see above). This expression also assumes you've appended the study dataframe as d in the results list in the .Rdata file you imported.


This page provided just the briefest of overviews of how you can use knitr/Sweave/xtable in your TeX document. There's much more documentation available that can help you do more complicated things with these tools. Here are some resources: