On Sun, Feb 22, 2009 at 5:47 AM, LuKreme <[email protected]> wrote: > So, I am 95% certain that the true probability is in the range of > 16.72% to 35.77% or so. So, my 'sense' that 19/78 is pretty close to > 25% is pretty much shit.
The spread isn't that bad. If your suspicion is correct (and it might well be) that the actual "rate" is 1 in 4. This test tells you that both "1 in 5" and "1 in 3" must be close to being outliers (given a normal distribution about your value of 19/78). If you want to ask the question "how sure am I that the rate is 1/4", have the test repeated a couple of times (with n≥60), and use regression analysis to compare 4 fits of y=mx+c for all populations 1. best fit 2. y=3x 3. y=4x 4. y=5x where: y= number of observations x= sum of successful outcomes m= "1/success rate" c= 0 (assumption - but you could, for extra robustness, leave this open in the best fit to test the strength of the assumption that c=0) If you find an insignificant likelihood that the value of m in the best fit is significantly different from 4, but a significant likelihood that m (best fit) is significantly less than 5, and more than 3, then you'd have what I'd call a useful answer. _______________________________________________ OSX-Nutters mailing list | [email protected] http://lists.tit-wank.com/mailman/listinfo/osx-nutters List hosted at http://cat5.org/
