Getting started with data
- Before even collecting your data, make sure your spreadsheet is set up in a way that ensures easy data collection and that minimizes errors in data entry. Be aware of silly things that happen when using excel, like the conversion of numbers to dates. Check out this blog post for some specific suggestions.
- Take detailed notes of all of your steps, and do as much of the data cleaning in R (or julia or whatever, but use code) to have reproducible cleaning steps and to preserve original datasets. Read Kate Laskowsky’s amazingly detailed suggestions and/or watch her youtube video to learn more.
- PLOT FIRST! Always. It doesn’t have to be a beautiful plot, but it should be the right type of plot to visualise the type of data that you have. Plot as much of the raw data as you can, and show the variation.
- Save your data in multiple places! Back it up!
- Go back to the basics from your stats class to consider how to analyze your data: probability distributions. The distribution of your data will influence the analysis methods.
Leveling up with github
- I find that I usually set up a github repository after I’ve already starting writing some code for a project (or at least already have some documents in a project folder, like data). So I often return to the helpful instructions on how to initialize an existing repository on the command line and on GitHub Desktop
- Add and Commit many small changes (and/or do this frequently) and use really helpful, informative messages about what you actually did in the change using
git add -p; git commit -m "informative message". Or type in a longer message in the text editor with just typing git commit after you’ve added your change. In general, add and commit at the same time to make sure the informative messages are linked to specific changes! - You can set your configuration settings to convert https urls to ssh, allowing for easy and pain-free switching between GitHub Desktop and command-line or CLI github. All it takes is one line of code! Thanks to @cmeyerton for sharing the one-liner solution on github forums
