From: Bobbi
Date: 02 Dec 2003
Time: 11:04:58
Does anyone out there know of a specific reference for justifying the maximum allowable error? I'm using the E/s ratio (maximum allowable error / sample std. dev.) to calculate the required sample size at alpha = 0.05. The application is fairly high-risk, so I expect the sample size to be large. Industry practice seems to be to estimate the error I'm willing to accept, and calculate the sample size from there. But aren't there some published guidelines for how to select the maximum allowable error? I will have to carefully justify whatever I select.