--- title: "Using table1 with \\LaTeX{} and MS Word" author: "Benjamin Rich" date: "`r Sys.Date()`" output: pdf_document: #latex_engine: xelatex #latex_engine: lualatex #keep_tex: true tables: true header-includes: - \usepackage{booktabs} - \usepackage{longtable} - \usepackage{array} - \usepackage{multirow} - \usepackage{wrapfig} - \usepackage{float} - \usepackage{colortbl} - \usepackage{pdflscape} - \usepackage{tabu} - \usepackage{threeparttable} - \usepackage{threeparttablex} - \usepackage[normalem]{ulem} - \usepackage{makecell} - \usepackage{xcolor} vignette: > %\VignetteIndexEntry{Using table1 with \LaTex{}} %\VignetteEngine{knitr::rmarkdown} %\VignetteDepends{kableExtra,flextable,printr} %\VignetteEncoding{UTF-8} --- ```{r echo=FALSE, message=FALSE, warning=FALSE, results='hide'} library(kableExtra, quietly=TRUE) library(flextable, quietly=TRUE) library(table1, quietly=TRUE) try(detach('package:printr', unload = TRUE), silent=TRUE) # Make sure printr is not loaded f <- function(x, n, ...) factor(sample(x, n, replace=TRUE, ...), levels=x) set.seed(427) n <- 146 dat <- data.frame(id=1:n) dat$treat <- f(c("Placebo", "Treated"), n, prob=c(1, 2)) # 2:1 randomization dat$age <- sample(18:65, n, replace=TRUE) dat$sex <- f(c("Female", "Male"), n, prob=c(.6, .4)) # 60% female dat$wt <- round(exp(rnorm(n, log(70), 0.23)), 1) # Add some missing data dat$wt[sample.int(n, 5)] <- NA label(dat$age) <- "Age" label(dat$sex) <- "Sex" label(dat$wt) <- "Weight" label(dat$treat) <- "Treatment Group" units(dat$age) <- "years" units(dat$wt) <- "kg" ``` ## Introduction A long requested feature has been the ability to use `table1` in `rmarkdown` documents that render to \LaTeX{} or `.docx` (i.e. Microsoft\textsuperscript{\textregistered} Word). Since version 1.4 of `table1`, this is now possible (with some limitations) by converting the output of `table1()` to a `data.frame`, `kableExtra` or `flextable`, using the functions `as.data.frame()`, `t1kable()` and `t1flex()` respectively, as these objects can be rendered to \LaTeX{} (note: `data.frame` (via pandoc) and `flextable` objects can also be rendered to .docx format, while, `kableExtra` cannot). ## Examples We demonstrate this using a familiar example from the main vignette. First, we can try the `data.frame` approach: ```{r} x <- table1(~ age + sex + wt | treat, data=dat) as.data.frame(x) ``` By default, this does not produce a \LaTeX{} table, but the same text output you would see in the R console. If the `printr` package is loaded, however, we do get a \LaTeX{} table by default: ```{r} library(printr, quietly=TRUE) as.data.frame(x) ``` Alternatively, we can use the `knitr::kable()` function: ```{r} kable(as.data.frame(x), booktabs=TRUE) ``` The output here is a bit nicer because we have specified the `booktabs` option, but because we are talking about a simple `data.frame`, there is no option to specify formatting (like bold text for variable labels, for instance). Next, we can try the `t1kable()` function, to produce a `kableExtra` object: ```{r} t1kable(x) ``` This looks a bit better: it uses the `booktabs` option by default, and also has bold variable labels. But because of a limitation with multiline headers (i.e. headers that contain line breaks), the `N=XX` for each column is placed in the first row of the table, which is less than ideal. Finally, we can try using the `t1flex()` function to produce a `flextable` object: ```{r} t1flex(x) ``` (Note that that `flextable` output, in particular the font, is different depending on whether the `xelatex`, `lualatex` or `pdflatex` engine is used, and `flextable` emits a warning when `pdflatex`, the default for `rmarkdown` documents, is used because `xelatex` and `lualatex` gives more font options.) The output is less attractive in my opinion, and less consistent with the typical look of \LaTeX{} tables and documents. It is more spaced out, doesn't have bold labels or line break in the column headers (actually, this is a problem common to both packages, but the `t1kable()` and `t1flex()` functions deal with it differently because `flextable` will actually include the line breaks in HTML and .docx output, and the overall best approach is unclear), and doesn't use the `booktabs` package (the thickness of the horizontal rules is different). Also, `flextable` places the table in a float, whereas `kableExtra` leaves it inline. I personally would use `kableExtra` over `flextable` in a \LaTeX{} documents, and in fact I have made this the default output in a \LaTeX{} context (i.e., when using `rmarkdown` with a \LaTeX{} output format). But a big advantage of `flextable` is that in can render to .docx format (i.e. Microsoft Word), which `kableExtra` cannot, and is the default output in that context. ### Nested groups Nested groups are supported with `kableExtra` and `flextable`, but not simple `data.frame`s. Here, one example with `kableExtra`: ```{r} x2 <- table1(~ age + wt | treat*sex, data=dat, overall=FALSE) t1kable(x2) ``` ### Captions and footnotes Captions and footnotes are supported with both `kableExtra` and `flextable`. An example with `kableExtra`: ```{r} x <- table1(~ age + sex + wt | treat, data=dat, caption="Test caption", footnote="Test footnote") t1kable(x) ``` Alternatively, the `kableExtra::footnote()` function can be used to add footnotes to the resulting object (this function also does automatic numbering, which may be considered an advantage). Note that when a caption is used, the table is rendered as a float. ## Closing remarks As of version 1.4, it is now possible to use `table1` within \LaTeX{} documents. There are some limitations on the formatting, however (it looks fine, but not quite the same as the HTML output and there is less flexibility to control it). The actual \LaTeX{} code generation is handled by an external package (i.e., `knitr`, `kableExtra` or `flextable`). The \LaTeX{} generated by `kableExtra` and `flextable` is very different, and in my opinion `kableExtra` produces the better result. Both `kableExtra` and `flextable` have an extensive number of functions that can be applied to the converted object in order to alter the appearance of the table but these will not be described here; refer to each package's documentation for the complete details.