-1

Workflow for statistical analysis and report writing

This question had a lot of good answers, but as pointed out, they are outdated.

I mostly work on scripts that will probably never be re-run after a paper has been published. Are packages worth the trouble in cases where I don't need to redistribute the codes to the world for easy access? What about the organization of data? How can makefiles be used?

Community
  • 1
  • 1
maximusdooku
  • 5,242
  • 10
  • 54
  • 94

1 Answers1

1

I think if you use the basics laid out by Josh Reichs in that post you provided, making sure that you create a directory to save everything in, then you are good to go.

My added step for the modern world would be to product a markdown report in one of the available formats.

  • rMarkdown- which you can run right out of rStudio
  • rNotebooks - which you can run right out of rStudio
  • Jupyter Notebooks - which you can run out of Anaconda or Jupyter with some easy tweaking.

The beauty of these three report systems is that you get to integrate the thought process, code, data, graphs and visualizations in a single spot.

So, if as you say no one will ever re-run your code, then they will at least see it to appease suspicions. Also, if they do choose to repeat your process, they just follow your logic and process in a duplicate document (especially easy with the notebooks)

As for using packages. That is a more complex question. If the packages are well orchestrated and save you a ton of time cleaning, sorting and structuring data, USE THEM! Time is money. If the things you are using them for are simple, straight forward, just as easy to program yourself and recognizable by those who would jury your paper, it probably does not matter either way.

The one place where I feel it matters is complex processes that are difficult (read that as easy to do wrong yourself) and have been implemented, tested and vetted by prior researchers.

Using those packages garners credibility and makes it easier for peers to accept your methods at face value. But if you are on the cutting edge..you should feel free to slice away. Maybe make a package of your own!

sconfluentus
  • 4,693
  • 1
  • 21
  • 40