To be as open as we can about what analyses we did, how we handled the data, etc., I think we should make sure everything we do has a notebook available in the analytics repo with comments, etc.
Our process has been:
Explore data via jupyter notebooks
When we find cool stuff, discuss it and decide if it's worth including in our final presentation/product/demo
Then we dump it into an intermediate table for the API to access without having to do realtime queries
Basically, between steps 2 and 3, we've selected a notebook or query that produces the exact data we will put into the intermediate table. This ticket is just to make sure all of those notebooks/queries are stored in the analytics repo with comments.
[ ] Create a list of all the pieces of aggregate data we're presenting
[ ] Add all of those as checkboxes in this first ticket
[ ] Make sure every one of those pieces of data have a notebook associated with them (doesn't have to be 1-to-1). While working on this, we should check any assumptions we've made and either justify them or at least be clear about them.
[ ] Make sure the data we have in the table actually matches this output (either by forcibly rewriting it, or some other way)
[ ] Document this association so that interested folks can find it. I'd suggest:
To be as open as we can about what analyses we did, how we handled the data, etc., I think we should make sure everything we do has a notebook available in the analytics repo with comments, etc.
Our process has been:
Basically, between steps 2 and 3, we've selected a notebook or query that produces the exact data we will put into the intermediate table. This ticket is just to make sure all of those notebooks/queries are stored in the analytics repo with comments.