- #1
thomas49th
- 655
- 0
Say I have a a log-normal distrubution of data. I want to use the central limit theorem to calculate how big the sample number should be. I would use the geometric standard deviation and we're dealing with a log normal distribution, correct?
Using the CLT I can arrive at the equation:
[tex]n = (\frac{z.\sigma}{EBM})^{2}[/tex]
Where z is the corrisponding z score from the cofindence level. Using 95% confidence in a 2 tail test yields z to be 1.96. Sigma is the mean (it shouldn't matter if I use the geometric and oppose to the arithmetic right?) and EBM is the error bound for a population mean. Now I need some help here please, on what EBM should typically be? I'm still not sure what EBM actually is... is it the same as 'relative error'
Is this equation the right way about going to calculate the sample number required?
Using the CLT I can arrive at the equation:
[tex]n = (\frac{z.\sigma}{EBM})^{2}[/tex]
Where z is the corrisponding z score from the cofindence level. Using 95% confidence in a 2 tail test yields z to be 1.96. Sigma is the mean (it shouldn't matter if I use the geometric and oppose to the arithmetic right?) and EBM is the error bound for a population mean. Now I need some help here please, on what EBM should typically be? I'm still not sure what EBM actually is... is it the same as 'relative error'
Is this equation the right way about going to calculate the sample number required?