Counterfactuals Do Not Exist and Yet Physical Randomization … Randomizes Them?
It’s not just covariates that get distributed equally between treatment and control — it’s the counterfactuals too
Today’s substack is about a few things. It’s about the beliefs people have about randomization and the reason that it sits atop the hierarchy of causal inference. It has to do with how they articulate what randomization does, the words they use, the concepts they reference, and how all of that while true is nonetheless different from my own. And it’s about how randomization not only distributes covariates equally across treatment and control — it also distributes potential outcomes equally, half of which don’t even exist.
The Obvious: Randomization Distributes Real Covariates Equally Across Treatment and Control
Some came to causal inference after having been trained in or at least heavily exposed to randomized controlled trials. They may have a medical background or they may be in tech. But regardless their main exposure to causal inference was f…
Keep reading with a 7-day free trial
Subscribe to Scott's Substack to keep reading this post and get 7 days of free access to the full post archives.