Dennis, thanks for responding. I would guess the "error" would be comprised of the inaccuracy of the sample as far as determining our actual collection rate.
As to your other point, the size of the error IS probably more important than the percent of occurrence. Would that make the "variables sampling" formula more relevant as to determine sample size? Or is "variables sampling" for a different type of analysis than the one we are doing? As for the tolerable error, we'd like to play with that number to see how it affects the sample size. Again, the population is 300,000. A sample of more than about 1,000 would probably be prohibitive for us to thoroughly examine. I'm more familiar with "attribute sampling." I'd rather just plug in (for example): Population size 300,000 Confidence level 95% Percent of occurrence 25% desired precision 95% and kick out a sample size. I'm not as familiar with the "variables sampling" formula, as I'm not sure if the "sampling error" and "estimated standard deviation" apply to what we're trying to do. However, if "variables sampling" is the correct method, then we'll go that way. [EMAIL PROTECTED] (Dennis Roberts) wrote in message news:<[EMAIL PROTECTED]>... > First, I would ask: what are you defining as an ERROR in this case? > > Secondly, I would then suggest two things for you to consider: > > 1. is percent of occurrence REALLY as important as HOW LARGE errors might > be? if the % happens to be highER but, the amount of errors (say) is small, > it doesn't really amount to much. > > 2. WHAT size of an error are you willing to tolerate? THAT size helps to > determine what n you might need to accomplish your goals > . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
