If you consider US political polling, one key correction is demographics of people who respond well to optional polls. Of course, they have eventual ground trut

If you consider US political polling, one key correction is demographics of people who respond well to optional polls. Of course, they have eventual ground truth and are only predicting a 1bit outcome. Still, do you guys have any thoughts on corrections on the data you get in?