However, 3) Different designs actually might estimate different things ("estimands"); their interpretation and generalizability, & how plausibly they identify a causal effect, might not be the same! We clarify which is doing what & their trade-offs between internal vs external validity 4/n
