lcrawley
> Can someone help me figure out how to best use "predlog".
> The help page
> says it" estimates, but does not display, the regression
> implied by the
> varlist before calculating the predictions." If I need to
> calculate and
> display percentage change after transforming coefficients
> back from log
> scale, will that command help?
-predlog- was written by Rich Goldstein (STB-29).
I wrote on -predlog- on 30 Sept 2003. I'll
repeat first some of my comments there which might well
apply here:
=================
... if you want this, it is arguable
that you are better off with
. glm response covariate, link(log)
. predict predicted
There's a large literature in this field. The papers
of W.G. Manning in health economics are one valuable
resource. He is, or at least was, a member of Statalist.
=================
As for the question, percent change from what to what?
I am not clear what you want to calculate
As for suppressing the regression output, the
central idea of -predlog- can be taken out of the
code and modernised thus:
gen log<response> = log(<response>)
regress log<response> <covariates>
scalar rmse = e(rmse)
predict yhatraw
predict resid, res
gen resid2 = exp(resid)
su resid2, meanonly
scalar meanres = r(mean)
gen YHATRAW = exp(yhatraw)
la var YHATRAW "Straight Retransformation"
gen YHTNAIVE = exp(yhatraw+(rmse^2/2))
la var YHTNAIVE "Naive Retransformation"
gen YHTSMEAR = YHATRAW * meanres
la var YHTSMEAR "Duan's Smearing Retransformation"
Taking similar code and putting it into
a do file will let you see the output
you want to see. Evidently you need to plug
in your own names for <response> and <covariates>.
Nick
[email protected]
*
* 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/
