On Fri, 25 Aug 2000 12:00:03 -0400, "William Levine"
<[EMAIL PROTECTED]> wrote:
> I am teaching an Introduction to Statistics course in psychology, and in
> class the other day, I brought up the issue that SAT scores and IQ scores
> may not really be interval scales. My TA pointed out that they are
> *standardized* tests, and that the intervals really are designed to mean
> something. I came back with the comment that we really don't know what these
> tests measure.
>
> Without raising the specter of what IQ and SAT tests really measure, does
> anyone have any comments that might be helpful in telling students what to
> think about this? Either that, or any references that discuss this issue
> explicitly.
It sounds like you have been misled into thinking,
"a scale is interval if ?
- I like its units, or something" instead of
"a scale can be equal-interval for a given purpose, and
a few will be equal-interval for many purposes
(but no scale will be equal-interval for all purposes)".
The question is whether you get linearity in your prediction.
If you are not trying to predict something, 'scaling' is a
misleading question.
For a couple of classical examples.
Distance is NOT equal-interval when you examine, say, the spread of
contagious diseases; the first 10 feet away from the disease is much
more important that 10-feet added to a mile or so.
Temperature (from -273 C.) is equal interval for many physical
processes, but it is not even monotonic as a measure of human comfort,
except within particular ranges. (Those might be useful ranges, and
it might be convenient to treat scores as "linear" in a study - but
you have to be given a particular range, in order to justify it.)
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
