View Single Post
Old 07-13-2010, 08:33 AM   #83
jbjb
Somewhat clueless
jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.jbjb ought to be getting tired of karma fortunes by now.
 
Posts: 779
Karma: 10535853
Join Date: Nov 2008
Location: UK
Device: Kindle Oasis
Quote:
Originally Posted by tompe View Post
Yes, of course. But depending on what you are studying and the result a smaller sample size can give statistical significant result. I read the comments about sample size in this thread as a disbelief that a small sample size ever could be statistical significant. And since that is a common error to believe that I just wanted to point that out.
That's exactly right - the number of samples required for a given level of statistical significance depends very much on the nature of what is being studied, the nature of its statistical distribution etc.

For example, if we *know*, a priori, that a die has a heavy bias towards a particular number, such that a roll of that die is known to have a 99% chance of being X, where X is an unkown in the range 1 to 6, then we don't have to make many rolls at all to have a very highly statistically significant idea of X.

(Please note that I'm not saying anything about the study in question, just following up on the more general statistical question.)

/JB
jbjb is offline   Reply With Quote