I ending up find the answer in Wooldridge,2002,p.699. If the variable
is not transformed it is a semielasticity, which makes sense as one
unit change in the variable causes a % change in the survival or
hazard ratio. If a variable is log transformed it is considered an
elasticity. In my case log(t)=log(-xb)+e for log-logistic.
On Tue, Mar 12, 2013 at 6:50 PM, James Henson <xenocobra1984@gmail.com> wrote:
> I have the basic understanding of the log-logistic (AFT) model estimates. If the
> coefficient is greater than one the it reduces the survival time, and
> increases survival time for values less than one. I am a little confused on
> if I use a variable such as the log of income (ln(income)) in a level-log
> format. Is it the same as regression (B/100)% change in beta. so, a one
> percent change in income would equal a (B/100) change in survival? Thanks I
> was search for days, but was unable to find the right answer.
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
