Some things I learned today

Stan files; R Markdown figures in LaTeX; Beamer in R Markdown; Build All and makefiles

Today (or, at least, recently) I learned:

Stan

You can open up a .stan file of Stan code in R Studio, and the syntax gets nicely highlighted for you. Also, there is a Check button, which checks that your file is syntactically correct before you try to compile it.

In addition, I finally got my head around (fixed effects) ANOVA Bayesian-style, but that is a blog post for another time. (From a Bayesian point of view, random effects is a lot more natural, but my students won’t have seen that yet.)

R Markdown figures in LaTeX

Yihui Xie is a genius (well, that much we already knew). I was generating a Beamer presentation from R Markdown, and the plots came out a bit fuzzy. I learned from The Man Himself in the first answer here that this is because knitr turns plots into .png images by default, and LaTeX doesn’t play nicely with those, especially if you want to resize them (which I did, because text-plus-plot on a slide makes the plot run off the bottom). The solution is to get knitr to generate .pdf plots, which you can do by putting a line like this in your setup chunk:

knitr::opts_chunk$set(dev = 'pdf')

Now my plots all look as if they have just come out of the washing machine.

I had previously thought that I would have to keep using LaTeX to keep using Beamer, but not so: the aforementioned genius has added Beamer support to knitr. What you do is to add something like this:

output: 
  beamer_presentation:
    latex_engine: lualatex
    slide_level: 2
    df_print: kable
    theme: "AnnArbor"
    colortheme: "dove"

to your YAML front matter, tweaking the options to your preference. (I don’t seem to have decided whether I am using lualatex or xelatex: see below. I don’t want to change anything because it is working.)

Build All

My lecture notes that I am working on (actually for two courses at once, but that is another story) are R Markdown files that have other R Markdown files as child documents, and the whole thing then gets run through LaTeX to produce a Beamer presentation. Knitting the root document (usually) works, but I wanted a bit more control over it, so I wrote a Makefile to automate the process. The Makefile looks something like this:

all: slides.pdf
    evince slides.pdf&
slides.md: slides.Rmd <lots of other Rmd files>
    Rscript -e 'knitr::knit("slides.Rmd")'
slides.tex: slides.md 
    /usr/lib/rstudio/bin/pandoc/pandoc +RTS -K512m -RTS slides_c32.md --to beamer --from markdown+autolink_bare_uris+ascii_identifiers+tex_math_single_backslash\
            --output slides.tex --slide-level 2 --variable theme=AnnArbor --variable colortheme=dove --highlight-style tango\
            --pdf-engine xelatex --self-contained 
slides.pdf: slides.tex
    xelatex slides.tex

The pandoc line was shamelessly cribbed from what appears when you click the Knit button.

Anyway, the reason for telling you this is that if you have a Makefile with an all target in it, the Build pane top right will contain a clickable Build All, and that will run your Makefile and build the All target. Likewise, there is a clickable More that will make the clean target if you happen to have one. I discovered this by accident, having discovered that I had a Build All in one of my projects, and I clicked it to see what it did. After a bit of rootling around, I found out (a) that this is what it was for and (b) I did indeed have a Makefile knocking around, that previously I had only ever run from the command line.