3 ms·

what do you do about sample bias?

• thrill 18y ago
  there's always sample bias - estimating and adjusting for it is the tricky part of course. The (statistical) "correction" for it is to insure the polling crosses all boundaries - and that's probably impossible - I know for example I have never particpated in a telephone poll - the interruption pisses me off for one thing - and I know many others like me in that regard. I have been called - and even answered some questions - but it rapidly became obvious to me that the wording of teh questions was tilted in some manner - so I declined to futher play that game. Being online certainly targets a certain audience - however the model discourages excessive influence (unless you've got deep pockets), and yet incurs such a small opportunity cost that anyone who desired can easily participate. My untested hypothesis is that anyone who makes an effort (that is minimal but more major than simply clicking) to participate when there is an opportunity cost will likely actually go vote for real - those are whom I actually want to sample. The only true test will be to compare after the election.
  • hugh 18y ago
    Am I reading this correctly? I have to pay you a dollar to participate in your poll? Good luck, but I think you're dreaming if you think you'll get many people to fall for that one.
    • thrill 18y ago
      Well, you're probably correct - while the top 10 political action committees spent about $120 million, gathered from a relatively small donor list, in 2004 to get their version of truth made into law, asking actual voters to spend an entire dollar as some sort of opportunity cost, with the thought that it might be the event that promotes an honesty in online polling, is quite likely simply a bridge too far.