Rich Ulrich wrote (in part)
<<<
Finally, some people like arbitrary transformations, including
adding arbitrary constants before taking the log or power:
What I am thinking of are the ones with the single virtue
of giving residuals that are apparently normal, for the data
on hand -- That is done in order to improve (or justify) using
the F-test.  The proper p-level is not achieved if you do not 
meet the assumption about residuals, so this DOES THAT.
I can admit that I did that a time or two, a long time ago,
and I might someday do it again.  
However, the F-test will be more simply wrong, if, say,
the linearity is fouled up by the transformation, making the 
coefficients wrong and mis-measuring the error.  I don't know
if I avoid 'arbitrary transformations'  because of that, or because
they are inelegant and hard to justify to anyone else.
>>>

I think an additonal point for (or against) such arbitrary
transformations is whether the original scale has any intrnisic meaning,
or whether the transformation does.  If the DV is some scale score,
where the scale is largely unknown and arbitrary, then it is less
'inelegant' and easier to justify than a similar transformation where
the original scale is not arbitrary, or is well known.

e.g. If the DV is (say) weight, then there may be substantive
justification for taking logs, or for using the original variable; I
would be more hesitant to use a BoxCox transform, largely because it's
harder to interpret.

OTOH, if the DV is some score on some test that no one's ever heard of,
then the original scale is totally arbitrary, so transforming it doesn't
make it MORE arbitrary.

(Of course, all this assumes that the transformation solves more
problems than it causes.....)


Peter

Peter L. Flom, PhD
Assistant Director, Statistics and Data Analysis Core
Center for Drug Use and HIV Research
National Development and Research Institutes
71 W. 23rd St
www.peterflom.com
New York, NY 10010
(212) 845-4485 (voice)
(917) 438-0894 (fax)


.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to