@OHBMequinoX 1. We joke about how terrible academic code often is. But as science becomes more dependant on code, it starts to scare me. Here are 7 tips for reducing the chances of having to retract your paper due to bugs in your analysis scripts. #OHBMx 
@OHBMequinoX 2. Divide and conquer to reduce complexity. Tip 1: cut up that 1000+ lines monstrosity of a script. Tip 2: have a strict division of responsibility between individual steps of your pipeline. Tip 3: make it easy to run a single step on a single subject, as well as run all. #OHBMx
@OHBMequinoX 3. Never trust someone else's code. Never *ever* trust your own code. Tip 4: Save the intermediate result of each script. Tip 5: Plot everything! Just because the end result looks reasonable does not mean you didn't screw up somewhere in the middle. users.aalto.fi/~vanvlm1/conpy… #OHBMx
@OHBMequinoX 4. Tip 6: Please, please, please, stop copy/pasting everything. Have a single config file for all parameters. Here is something revolutionary: define your filenames *once*! Have templates for them! This will change your life. #OHBMx
@OHBMequinoX 5. Tip 7: stop hoarding files!! Apply konmari.com to your code folder. Put everything in version-control (git) and then *delete things*. I've seen people dig for half an hour to find the one script that constituted the "working" version.
@OHBMequinoX 6. Let your code be a testament to your scientific rigor and technological creativity. Read more about the why and how of these 7 tips at arxiv.org/abs/1904.06163.
Do you have any tips to share? Reply to this thread and let's discuss! #OHBMx
Do you have any tips to share? Reply to this thread and let's discuss! #OHBMx
@OHBMequinoX Important note to this tip: please do put everything into version control and back it up. Never really irrevocably delete anything human-written! Stuff it in the endless depths of your git history, push to somewhere safe.
@OHBMequinoX Just realized there is a bug in my code example above. Because of course there is :) All code is terrible.
