Negative selection and instrumental variables
Highly active antiretroviral treatments, HIV and AIDS, and American healthcare quirks
Selection bias
As I work on the revision to the Mixtape, I am finding myself remembering with fresh eyes old papers I haven’t thought of in years. What is selection bias exactly? To many of those who don’t obsess over causality, selection bias is associated with non representative samples of data. For instance, if I interview the first 100 people I meet at the coffeeshop today and ask them about their spending, I can go home and run regressions relating income to individual characteristics, but I really shouldn’t try to make claims that whatever I find in that data is true for Waco Texas, the city where I live, because my sample is likely biased. Thus for many people, selection bias is a sampling concept. Had I randomly sampled Waco citizens, instead of just conveniently sampled them, then those associates in my data might actually conform to patterns for the city at large, but not otherwise except by the wildest of coincidences.
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.