This repository contains the files for the presentation given by Alex Dolphin at the Düsseldorf R User Group Meetup on 17/10/2019
Slides from Alex’s talk "Practical reproducibility of analyses"
In the pursuit of robust, reproducible, and extendable analyses, one key aspect frequently gets overlooked: practicality. Reproducible analyses with a potential for good review are often cumbersome to write: they may require branches to be pulled, environments to be built, and variables to be changed in a well-hidden file. I will propose a step towards greater practicality with an example using three technologies many data scientists know to some extent:
- RMarkdown
- R shiny server
- Docker
By presenting this concept I hope to demonstrate that it is possible for:
- data scientists to review the live code with one link to a shiny-rendered page
- non-data-scientists to rerun and extend the analysis through an intuitive GUI
- the data scientist doing the task to achieve this only writing one file in their preferred coding language
The presentation contains 4 links. Unfortunately:
- One of these refers to the shiny server running on my local machine, and will not work
- The other three refer to locations inside the trivago network, and unfortunately will not work for those not connected to the network.
Two of these links are trivago-internal repositories, and unfortunately the code cannot be made public at this time.
As the presentation doesn't offer much practical information without the repositories, you can get in touch with me, the presenter, either via:
I will be glad to discuss the concept, or general R/data/data science topics.