ColinsMum, I understand what you are saying - the "small child" type variables apply just as much to the sample group as to the individual child.
What I don't really understand is how the example of regression to the mean that Cathy posted applies to something like IQ testing. A multiple choice test sure, particularly one that where everyone answers randomly (which isn't really a test at all!). But an exam that requires knowing the material and being literate, or an IQ test, while I can see there being variability from day to day based on how good you are feeling, raport with the tester, etc, it still seems to me that these are tests based on skill/ability and that you aren't going to randomly swing from the 99.9th to the 50th? Maybe 5% but not 50%!
You can do well on a multiple choice by "getting lucky", but I don't see how you can do well on tests like Vocab, Block Design, Word Reasoning, etc by luck. "Having a bad day" aside I am just not understanding the idea that children tested young will swing back to the mean, or why this is only thought to apply to young children.
Under performing because of being young makes perfect sense, over performing really doesn't. So surely young children are more likely to trend up than other test subjects?
