Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="kenobi65" data-source="post: 3555803" data-attributes="member: 1515">(Stats hat on; I do market research for a living)There's something called the "Law of Large Numbers"; when your sample gets big enough, you can feel fairly confident that the results you see in the sample are going to be pretty similar (though not necessarily identical) to the results you'd see if you did a census (that is, if you got an answer from everyone in the population)*.Generally speaking, once your sample size gets to the 200 to 300 range, you don't see a whole lot more precision from adding additional responses.  OTOH, results from a sample that's substantially under 100 (as this poll currently has) are fairly likely to differ substantially from the result you'd get in a census.And, actually, "sampling bias" may not have been the best choice of words here.  What I talked about above is just an issue of a small sample size.  Technically, "sampling bias" is a situation in which your results may not be accurate / predictive, because you didn't get a good sample of the entire population.  For example, every time you see people here talk about "well, that can't be true of D&D players, because no one here on EN World does that!", that's sampling bias.  EN World isn't a represenative sample of D&D players, and thus, using what EN Worlders say, and assuming you can project that out to the broader population of "all D&D players", is a flawed assumption.* - Assuming that your sample is representative.  Polls in which people choose whether or not to answer are notoriously un-representative, because there's no way to tell whether the people who seek out the poll and answer it are the same or different from those who didn't see it, or chose not to answer it.</blockquote>

Verification