Quote:
Originally Posted by danmckee
If there are 300 or 500 green Cobbs out in the hobby, and Ted has survey'd 30, is that a good testing environment? Can you take the 30 and say the results cover the remaining hundreds that you didn't see? I say no.
|
I guess it depends on the definition of good testing environment. Statistics are used all the time using small sample sizes to extrapolate to the actual population (polling estimates, Nielsen ratings, etc.). Statistically, given a universe of 500 and a sample size of 30 with a percentage of 87, you can be reasonably confident (95%) that the margin of error is +/- 12%. In other words, it is 95% probable that the actual percentage in the population is between 75% and 99%.
Plug in any numbers you want in the "Find Confidence Interval" calculator here:
http://www.surveysystem.com/sscalc.htm