Drake for KnoxData

This is a presentation by Brian O’Meara for the KnoxData group on March 19, 2020, on the drake package for making a workflow in R. Thank you to Josh Rosenberg & Alex Pawlowski for inviting me.

The talk is available here.

You can see code examples made for this at https://github.com/bomeara/drake_knoxdata; to download everything to your computer, https://github.com/bomeara/drake_knoxdata/archive/master.zip

This also uses

• tidyverse
• rmarkdown
• cowplot
• future

The example uses data downloaded from https://transtats.bts.gov/DL_SelectFields.asp: Flight arrivals and departures for US airports for December, 2019.

Learning objectives

• Utility of workflows
• How and why to use
  • drake
  • RMarkdown
  • Parametric RMarkdown

Outline of presentation

• Who am I: http://brianomeara.info. I study evolution by creating and using new methods, often implemented in R. You can see my github profile, too.
• Why drake?
  • Reproducibility: https://github.com/bomeara/ReturnOfTheMinivan analysis for a paper (https://doi.org/10.1002/ajb2.1278)
  • Scalability: Comparing over 1,000 faculty Google Scholar profiles for department 10 year report.
  • Saving time: Tweak something, only things that depend on it need to be updated
• Why RMarkdown?
  • Make reports that can easily be changed with different data
• More info: Free drake book (with videos, too): https://books.ropensci.org/drake/
• The old way: https://github.com/bomeara/drake_knoxdata/blob/master/basic_scripts/flight.R. We call the R packages we need, load the data, make the output.
• A basic drake way: https://github.com/bomeara/drake_knoxdata/blob/master/drake_flight_simple
  • In this directory, open R, then you can do source("make.R") to run the plan.
  • You’ll see plots being made (but the order might differ from the script only version)
  • vis_drake_graph(plan) to see how the plan works
• A more fully featured drake way: https://github.com/bomeara/drake_knoxdata/blob/master/drake_flight_complex
  • In this directory, open R, then you can do source("make.R") to run the plan.
  • You’ll see plots being made (but the order might differ from the script only version)
  • vis_drake_graph(plan) to see how the plan works
  • This uses parametric Rmarkdown
• Output from this: https://bomeara.github.io/drake_knoxdata/airport_overview.html