Quote:
Originally Posted by 3pete
I suspect you are thinking of surveys based on a random sample. Given the decidedly non-random approach taken for this survey, no such generalization can be made. Think "Family Feud" rather than Gallop polls.  |
Based on the questions asked, it was most likely supposed to be as close to a random sampling as Macmillan could get from their own mailing list. However, since the link was posted here, the results are now very skewed towards a non-representative sampling of readers. At the same time, though, they can filter the results based on the unique string in the URL. If they're smart about it, they can now have two data sets from which to draw conclusions--one comprised of a fairly random sampling of their mailing list subscribers, and another comprised of e-book enthusiasts.