Rich Ulrich <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>... > The Finite Population Correction (FPC) is not used in > standard political polls -- which only tap a tiny fraction > of the voting population. It is used on election eve, > when votes are coming in, and then it is used with great > caution; or else folks can end up predicting (say) that > Florida has gone to Bush by 100,000 votes.
Ouch, looks like I've been using the wrong formula for awhile, then. Could you give me a reference to the right way? ru: > Unfortunately, the theory is dull and tedious. I'm finding that out as I consult with others! ru: > If you have pilot data, you can compute the test-total, and you > can extrapolate to larger Ns and larger power. If you don't > have any numbers to start with, I don't think you can get very > far. I don't see much as a theoretical problem, and I don't > see detail that allows better referrals for the concrete problem. > - The general problem of group-dependencies, of course, > exist outside of 'regulation'. There are some references > concerning 'effective sample size' in a post by Jon Volstad, > saved in my stats-FAQ at > > http://www.pitt.edu/~wpilib/statfaq/96sampn.html Thanks for the link - he hints at the ability to correct for these 'design effects', but looks like I'll have to pursue the references to find out how. se: > > it seems to me it would come up in many instances of evaluating > > organizations as a whole, with many individuals, performing multiple > > tasks (e.g a factory, with many employees, making many widgets each > > and you wanted to estimate *factory-wide* the proportion of defects in > > the widgets that was occurring - this is the *exact* same problem that > > I have) > [ ... ] ru: > From this, it seems like someone wants a small-variance > estimator of an overall total. "Stratification of surveys" comes > to mind. Maybe someone has a formula or a reference, > but you still would need to have an estimate of the > dependency, measured as an intraclass correlation or > something similar. I'm not sure what you mean by a 'small-variance' estimator, but yes, I'm after a number that represents the total organization. This factory example I thought demonstrated the generic and far-reaching nature of this problem. I have to disagree with your statement above to the contrary (regarding it not containing a 'theoretical' problem - whatever that means). I'm not sure HOW these organizations do it, but my guess is that a CEO would want to compare the defect-preventing ability of different factories based on ONE overall estimate of the proportion of possible defects that OCCUR in each factory, and dammit, I'm sorry, he should be able to get it in an accurate fashion. Maybe in the 'real' world, they just assume independence and don't worry about it - my colleague said that that always makes analyses more conservative anyway, so the only danger is inflated Type II error. Oh well, I realize that probably the theory and math necessary is likely beyond me, but I simply can't believe that this type of problem hasn't come up in thousands of other situations. I'm getting convinced that the answer may lie in the multilevel modeling field, which embraces and accounts for dependency, but alas, when I was in grad school, it was just getting hot. Perhaps I need to do a little reading... Finally, regarding your comments about the nature of this group in general, it appears that you are saying that it's primary purpose is to point somewhat newbies to 'at worse a bit obscure' answers. This is a VERY valiant cause, and I'm sure that I could learn a thing or two from reading it, but if that is so, where would be a more appropriate place for a somewhat-seasoned data analyst to post questions of a more 'obscure' nature? I can only find 3 stat newsgroups, and they only other one that seems like it might be the one is sci.stat.consult. Thanks for your help, Rich. I appreciate the dynamic exchange. Scott . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
