On Wed, Mar 10, 2010 at 12:32 PM, Robyn Bergeron <robyn.bergeron@xxxxxxxxx> wrote: >> That looks to be about 400 people need to randomly selected and >> complete the survey (for +/- 5%). to get down to 1% you would need to >> get 6500 people. > > I don't think that the near-impossibility of having a statistically > sound sample that would hold up in a court of law means that we > shouldn't at least get a feel for who users are by doing a survey. > Certainly, an somewhat unstatistically-sound sampling is much better > than all of us guessing, is it not? :) > I am not looking for something to hold up in court.. however currently we have very contentious issues that people are wanting to survey. And the results will be used to 'settle' those issues in the minds of the 'winners'. [No less than when a partisan's candidate is shown to be ahead but still within margin of error, they will just state it proves that the masses are on their side.] And on the counterside, every thing that doesn't meet criteria will be brought as why it was rigged. To clarify, the surveys as you mentioned in your previous email and the data you are looking for are good 'first' steps that any series of surveys need. And you have been clear in the beginning that they aren't meant to clarify larger issues like: KDE vs GNOME vs LXDE, fast updates versus no updates, boxers versus thongs... they are mostly to get a general feel where it doesn't really matter if your error is +/-10% or that there is small amount of a self-selection bias. In the case where you wanted to start drilling down "Why do programmers prefer boxers but our QA people prefered thongs, and people who said they were just users said Hanes?" then you want to start being more rigorous. Does the above better clarify for you my point of view? -- Stephen J Smoogen. Ah, but a man's reach should exceed his grasp. Or what's a heaven for? -- Robert Browning -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
