Open jcolomb opened 8 years ago
I appreciate your point. I did have a "techie" audience in mind: folks who write a lot of code in R, but not always in a way that is reproducible.
For the non-programmer, your suggestions are important in moving towards improved practices, but I think for them to get to reproducibility, they should first learn to write some code.
I am clearly that audience an that is probably why I was so impressed by the slides. But from experience, I know this audience is quite limited. the first step is to convince people to learn scripts.
One has to keep in mind that pencil and paper is the most used tool for most researchers. (And this is also normal, in many cases, pencil and paper are the best tools...)
.
On 12 January 2016 15:30:55 CET, Karl Broman notifications@github.com wrote:
I appreciate your point. I did have a "techie" audience in mind: folks who write a lot of code in R, but not always in a way that is reproducible.
For the non-programmer, your suggestions are important in moving towards improved practices, but I think for them to get to reproducibility, they should first learn to write some code.
Reply to this email directly or view it on GitHub: https://github.com/kbroman/steps2rr/issues/5#issuecomment-170927717
Envoyé de mon téléphone Android avec K-9 Mail. Excusez la brièveté.
I think most people are not able to envisaged scripting at all. The way your lesson is build, you will loose these people very fast (on the first step actually). I would try to go more easy on that. I would go in 3 initial steps:
In more details, it is a lot about moving things around:
Step 1
Organise your data
1 project = 1 directory get the data in the most-raw form possible. Separate raw data from derived data Write ReadMe files/ get a master file explaining your organisation and describing the experiments
Choose file names carefully
Avoid using “final” in a file name. File names should be explanatory, they can be long Avoid spaces, use _ use read.txt files to explain your naming, if not in the master file
If this is done, we would already achieve a lot: data is easy to share, and scripting may be outsourced.
step 2
From there one may become more technical (from code commenting to the use of MAKE, github and figshare...)