Saying "Race is a risk factor" is problematic. Statisticians and social scientists need to be more careful about this kind of language and avoid it where possible ...
Race is a proxy for heredity which is a proxy for genes. Genes exist and are causal. While you can relate race to outcomes, it's not special! ANY arbitrary heritable partition of humans that correlated to genes could substitute for race. In this sense, Race is not "real"!
Imagine giving 10,000 randomly chosen people with sickle cell a ring. Tell them to pass it down father to son. Will you find that in 100 years people with this ring are likely to have sickle cell? Yes! Yet the property "has a ring" is in a deep sense completely arbitrary!
If ring families as much as possible only marry into other ring families, this pattern will persist over time. Because this is an observable pattern, it is detectable by the scientific method but it's not real in the sense of being a natural property of the universe.
Even worse than this, race is not causally well-defined. It involves at least three distinct mechanisms: 1. Heredity 2. The individuals perception of their own race 3. society's treatment of that individual due to race.
Studies that measure "race" are likely capturing completely random and extremely different mixtures of causal mechanisms. I don't say never use race. Just remember it's often a low quality substitute for the thing you *really* want to measure and admit this in your study!!!
Thanks for coming to my TED talk.
