On 6 Oct 2002 18:30:00 -0700, [EMAIL PROTECTED] (Paul Bernhardt) wrote: > It sounds like he's working with IQ test data which is normed to a mean > of 100 and SD of 15. If that is the case, he should first check that his > sample's variance is not different from the norm for the test, then use a > Z test. If it is different, use t-test. >
It has been my experience with psychological tests that the published norms can be reasonable for the means; but my own samples *very* often have a different SD. That is why I would not start with the assumption of having the same SD, especially if my N was anything reasonable - large enough for a halfway stable SD. There is a question here that I have never asked before. I would be willing to make a statement (of a sort) about a single case, using the "known SD=15." I would probably comment on "two individuals" the same way. When (what N) should I switch to the group-defined SD? Is there a 'Bayesian' solution that smooths the switchover? -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
