Assuming that you
-- have a dichotomous dependent variable
-- are using the conventional 95% confidence level
-- want the margin of error around an obtained percentage to be no larger than 1
percentage point.
-- don't have an expectation of what the resulting percentage will be
n= N * p * q / ( ((N-1) * (B**2) / (t**2) ) +
(p*q) )
n = 2600000 * .5 * .5 / ( ((2600000-1) * (.01**2) / (1.96**2) ) + (.5*.5) )
Scheaffer, R.L. Mendenhall, W, & Ott, L.(1990) "Elementary Survey
Sampling",Belmont, CA: Wadsworth. p 74 formula 4.19.
if you do have an expectation,e.g., that 20% of the cases will have the
characteristic you are interested in then use .2 for p and .8 for q.
Your standard deviation is sqrt((p*q)/(n-1))
Brian Vuong wrote:
> Dear stat. subscribers:
>
> I work for a bank. I have a data set with 2,600,000 observations. I would
> like to determine an auditing sample size that is statistically significant
> within 1% error. If I can determine the sample size I need, I will use SAS
> to pull out the sample size randomly, so I can do auditing. I went back and
> read my old statistics book, I found this formula with standard deviation,
> but I don't have any standard deviation, so I don't think it is a right
> formula. Thank you for your help.
>
> Thank you for your help,
>
> Brian Vuong
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
