[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Transformed values in logistic regression
On Thu, 26 Aug 2004, Ricardo Ovaldia wrote:
> Dear all,
> Please pardon this non-Stata question, but I have not
> found a satisfactory resolution.
> them into our logistic model. However the reviewer
> wrote ďUsing a log-transform for creatine is absurd
> because a 1-unit increase in ln(x) is equivalent to
> increasing x by a factor of 2.718 which is in the
> realm of impossibility
Independent variables only need to make sense across the actual range of
the data, lots of things can't be negative, or more than 100%, etc. The
objection for that reason doesn't make sense, unless at some point in the
paper you illustrate the effect of creatine with a one unit change. If
that is the case you may wish to substitute a one standard deviation
change for the illustrative change.
> Is he correct? Isnít the coefficient estimated such
> that the predicted values are within the range of the
> data and this only a problem if you attempt to
> extrapolate beyond the data range? What I am missing?
The logit transformation does ensure that the dependent variable is always
feasible, (log variaes across the entire real line, while probability is
constrained from zero to one). However the reviewer's complaint does not
seem to be with the dependent variable, but with one of your independent
However, before disputing the reviewer's claims, I would rerun the
regression without the log transformation. If the result is still
satisfactory, you can save a lot of touris by caving now. A wise man once
taught me that you shouldn't fight a referee unless his request would be
very bad for the paper.
> Thank you very much,
> Ricardo Ovaldia, MS
> Oklahoma City, OK
> Do you Yahoo!?
> New and Improved Yahoo! Mail - 100MB free storage!
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
* For searches and help try: