6 Reporting

After the successful processing and visualization of the data, the results need to be reported. This can be done best in a “literate programming” fashion as provided by R Markdown.

Using R Markdown, one is able to combine R code (and its results) with text (written in markdown) to create professional looking reports in various output formats (Word, PDF, HTML).

Both interactive and static documents can be created. This gallery (maintained by RStudio) gives a first overview of how documents created using R Markdown can look like.

6.1 Overview

File Format: .Rmd (R Markdown)

New document (in RStudio): File -> New File -> R Markdown/R Notebook. LaTeX Math is supported via Mathjax:

\(y=\frac{(x - \mu)}{(max - min)}\)


Any file with the .Rmd file extension is an “R Markdown document”. RMD’s consist code and text (written in markdown syntax) which need to be compiled into a high-level output format.

Possible output formats are:

  • HTML

  • PDF

  • Word

  • Powerpoint

Which output format should be used is specified in the “YAML header” of the R Markdown document.

6.2 The YAML header

In the YAML (Yet Another Markup Language) header users can specify metadata which denote the final appearance of the document.

Each output format has different settings. Fortunately, most settings apply to all formats.

The YAML header starts and ends with three dashes: ---. The output field is mandatory.

Valid options for each output format can usually be looked up in the help page of the specific output format.

For the default output format html_document the R Markdown - The definitive guide book is a good reference.

An R Markdown cheatsheet also exists.

6.3 Literate programming in R

Packages {rmarkdown} and {knitr} are the base of literate programming in R.

RMD’s documents can be compiled

  • by clicking the “knit” button in RStudio (the name relates to the {{knitr}} package)

  • via the command line by calling rmarkdown::render()

Behind the scenes the {rmarkdown} package first converts the .Rmd file to .md (markdown). Then pandoc, which is a universal markdown converter library, converts the .md file to the chosen output format.

6.3.1 R Markdown packages

The following packages are built upon {rmarkdown} and simplify special purposes.

  • bookdown: Mainly used for writing books but can also be used for reports (formats html_document2, git_book, pdf_book, etc.).

  • thesisdown: A package for thesis writing. Provides ready-to-go templates for different types and simplifies advanced LaTeX usage.

  • rmdformats, pinp : Different templates for literate programming documents.

  • xaringan: For HTML presentations via remark.js.

  • blogdown: For creating websites. Example: https://pat-s.me

  • rticles: For scientific paper writing in R.

6.3.2 Code chunks in R Markdown

To insert code into an R Markdown document, one needs to add a so called “code chunk”.

```{<language>}

```

This tells the document that everything within the three backticks should be interpreted as code using the given language. Code can be shown/hidden, evaluation can be prevented on demand, results can be cached, etc. See https://yihui.org/knitr/options/ for a full list of supported options

6.3.3 R Notebooks

R Notebooks are a special form of the html_document. To use it, specify html_notebook as the “output” type in the YAML header. This output format was created by RStudio as an alternative to html_document.

Differences compared to html_document:

  • Code output is shown inside the editor and not in the console (can be changed)

  • Instant preview of the output document without having to get all the code in the document running. The results from the last successful code execution will be used (if there was one).

  • Link in the HTML doc to download the source .Rmd file

  • Option to toggle on/off code chunks for the whole document

  • Output file extension is named .nb.html

6.3.4 Workflow

R Markdown documents are most often used for reporting of results created in an Rscript. This enables a seamless integration of data processing tasks into the subsequent reporting.

Reporting often splits up into different formats:

R objects (containing results) can directly be used in the reports to present the results (data, plots). If the complete workflow of an analysis has been set up in R, changes at certain stages of the workflow (e.g. incoming data) can easily be integrated.

This is the point where packages like drake, workflowr and rrtools jump in to simplify reproducible workflows in R.

A widely used concept is to start a project following the structure of an R package. This helps due to

  • a consistent directory structure of R scripts and R Markdown documents
  • documented custom functions
  • simplified integration into workflow packages like {drake} and friends.

R “research packages” can be installed locally like any other R package and simplify usage and sharing among colleagues.

6.4 Shiny: Interactive visualizations

Javascript based R ecosystem which provides options for rich visualizations. The shiny gallery from RStudio gives a good overview what can be done using shiny.