I forget my statistics...
The problem is that the variable is not binomial, it is within an assumed range (i.e. it is along the interval 0 to 25) in a pretty much infinite population. How should I select a sample size to determine mean within a reasonable confidence? It should be normally distributed.
The problem is that the variable is not binomial, it is within an assumed range (i.e. it is along the interval 0 to 25) in a pretty much infinite population. How should I select a sample size to determine mean within a reasonable confidence? It should be normally distributed.
-
The relationship is t s/sqrt(n) = B, where B is the margin of error required, t is the critical value value cutting off alpha/2 in the normal distribution, and s is the STD deviation.
Solving for n, n=(t/B)^2 x s^2.
For a 95% confidence, alpha=.05, and t=1.96.
the problem is often that you don't know what the STD dev s is.
Here, you know the range is 0 to 25, and you can use a rough estimate that the range constitutes 3 or 4 STD deviations, s=25/3 or s=25/4.
Solving for n, n=(t/B)^2 x s^2.
For a 95% confidence, alpha=.05, and t=1.96.
the problem is often that you don't know what the STD dev s is.
Here, you know the range is 0 to 25, and you can use a rough estimate that the range constitutes 3 or 4 STD deviations, s=25/3 or s=25/4.