Weights

[andreifoldes]

Nice to have columns in ones data release might be some type of weights e.g.:

• nonresponse weight: so we could use the screening survey variables to model the probability of signing up, then construct a var to up-weight those who did enroll but looked similar to the people who didn’t
• representativeness or calibration weight: align the (already nonresponse-adjusted?) sample to some external population margins (ie UK/US 18–40-year-olds?)