Stan, Thanks for your reply. We really do not want to score all 1600 if we don't have to. This local assessment will be a writing prompt. The students' writing will be assessed against two rubrics. One for writing and another for critical thinking. So, what I would like to know is, can we get an approximation for the population mean without having any baseline data, and if so, how many students would need to be sampled?
Thanks, Eric Lund Stan Brown <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>... > Eric Lund <[EMAIL PROTECTED]> wrote in sci.stat.edu: > >In Indiana, the school accountability legislation calls for locally > >produced assesments to be used. An assessment is to be given to about > >1600 students in the high school in which I teach. This assessment > >will be given each September and May starting with the 2003-2004 > >school year. How large of a sample must be drawn from this population > >in order to approximate the mean score for all of the students? > > The answer depends on three things: the population standard > deviation, the level of confidence you want in your estimate, and > the amount of spread you're willing to tolerate in the estimate. For > instance, if you can accept an estimate with a spread of plus or > minus 80 points at 95% confidence, you can take a smaller sample > than if you need a spread of no more than 20 points at 95% > confidence. > > However... > > If I understand your scenario, you don't really need to do any > sampling. You'll have the score of every student who took the test, > so you can compute the population mean score exactly. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
