3 ms·
Forgive me if this is a dumb question, but what is "voluntary response data"? Is the author arguing that the only statistically valid way to do a study is to pu
by j_baker 15y ago
Forgive me if this is a dumb question, but what is "voluntary response data"? Is the author arguing that the only statistically valid way to do a study is to put a gun to peoples' heads and demand that they take a survey? I can't think of many ethically sound ways to conduct a study without responses being voluntary.
- drewcrawford 15y agoThink about if the poll were (the horror!) political. The person who responds to that kind of poll is the intersection of: 1) An HN reader 2) Somebody who thinks politics on HN is a good idea 3) Somebody who is on HN / has free time TODAY, skewing the distribution in favor of certain verticals that are in a low point of their seasonal cycle or against startups in a coding marathon 4) Somebody who personally cares about politics sufficiently to have an opinion strong enough to want to be counted, this is likely to skew the distribution in favor of extremes In contrast, if pg picks 1000 random HN users and personally asks them the poll question via e-mail, some of these biases evaporate. And you lose the "interested bystander" effect whereby more interested people tend to want to take the poll, because the only way to take it is by personal invitation.