opensource-observer
Causal Analysis
What is it?
Generally speaking, we need a better way to think about how to show that funding OSO metrics has the intended effect, with retrofunding being the largest experiment to date.
To start with, we should have time-series rolling windows on all metrics so that we can show change pre/post funding. We should also be monitoring it over time, whether there is any stickiness.
In the future, we might consider controlled experiments to factor out other causes.
Also related:
- Relentless monetization
https://kermankohli.substack.com/p/arbitrums-85m-growth-campaign?r=30w0j0&utm_campaign=post&utm_medium=web&triedRedirect=true
There's lots of materials on causal inference:
- https://matheusfacure.github.io/python-causality-handbook/landing-page.html
- https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/
- https://aayushmnit.com/posts/2022-09-19-SyntheticControl/2022-09-19_SyntheticControl.html
Which is related to the broader field of econometrics
- https://www.aeaweb.org/conference/cont-ed/2017-webcasts
- https://www.aeaweb.org/conference/cont-ed/2020-webcasts
- https://www.mostlyharmlesseconometrics.com/
- https://www.masteringmetrics.com/
Starting some thoughts here https://docs.google.com/document/d/1S7jqzB_JkAyygZ7z3z6m5TVmBZWza7ngMlrrWASs0wA/edit
As per convo with @ccerv1 , he's going to cover this in his talk and in the blog post. I'm going to move this back to the previous sprint and close it out as the initial exploration