I think everyone can be flexible here according to the problem. I'd be
happier with working with the idea that Maarten is around the
99.somethinglarge% percentile of intelligence than with trying to
estimate his intelligence on any scale. If I am a pharmacologist (to
pick an arbitrary example) I trust the measurement scales my physical
and chemical forebears have spent centuries perfecting. I don't want to
throw them away as I think that a dose of something means what it says.
Nick
[email protected]
Maarten buis
--- Nick Cox <[email protected]> wrote:
> Roughly speaking, my experience is that natural scientists
> (physicists, biologists, etc.) are more likely to take normalised as
> meaning scaled to [0, 1], far more commonly by
> (value - min) / (max - min) than by percentile ranks. The more
> statistics you know, the more likely you are to regard
> (value - mean) / sd as a natural standardisation.
The funny thing is that (value - mean) / sd scores are often
interpreted in terms of percentile ranks: a value of -2 can usually be
considered small because, if the variable is reasonable close to a
Gaussian distribution, one would expect that approximately 2.5% of the
respondent would have a smaller value (and thus 97.5% of the
respondents to have a larger value). That looks suspiciously like a
percentile rank to me... This is one of the reasons why I think that
there is quite some merit in using percentile ranks as a form of
standardization.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
