Re: st: How to interpret an estimated model in log form in practice

["Dimitriy V. Masterov" <dvmaster@gmail.com>]

Talal,

This is a pretty vague question, but as someone who asks many such
questions, let me give it a try. In the future, it may help to provide
more detail about your problem, such as what the y and x variables are
and what estimation procedure you are using.

For some examples of useful geometric means in economics, take a look at:
http://davegiles.blogspot.com/2012/01/extracting-correct-meaning-from-data.html

For a slightly more relevant example to your problem, take a look at
the eform() option of regress described here:
http://www.stata-journal.com/sjpdf.html?articlenum=st0054

To transform your predictions back to the original scale, you need
levpredict from SSC. It's post-estimation command for use after a
linear regression model with a logarithmic dependent variable has been
estimated. It generates predictions of the levels of the dependent
variable for the estimation sample. These predictions reduce the
retransformation bias that arises when predictions of the log
dependent variable are exponentiated.

Finally, take a look at using glm with a log link function instead of
calculating ln(y) yourself. The assumptions about the error term are
different, but I've found it works very well and avoids this whole
retransformation business. I think glm can adjust for serial
correlation, but this getting into an area where I am not qualified to
offer solid advice.

HTH,
DVM
*
*   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/