Re: st: help log linear

[David Gottschlich <davidgottschlich@comcast.net>]

I seem to have gotten this a long time after you sent it, so you may 
have the answer already (and please forgive me if I giving the wrongly 
simple answer to a complex question), but....

predict logy
will give you the predicted result for each independent x; of course, 
this will be in log space.

If you used the natural log, then
generate predicty=exp(logy)
will convert the predictions back to real space

or
generate predicty=10^logy
works if you used the common log.

For most (many?) applications, it's more common to use the form
y=a ln(x) + b
than
ln(y)=a x + b

If you are using the second form, of course you cannot force y to equal 
zero unless you can reach negative infinity from your function.

David

cas2111@columbia.edu wrote:


> Hi,
>
> I am having a bit of difficulties figuring out how to predict a y
> in a log linear model.  I cannot determine the following items:
>
> 1)obtaining fitted values of logy(i) from the regression of logy
> on each independent x
> 2)for each observation i, creating a variable equal to exp(logy(i))
> 3)performing a regression through the origin
>
> Any other information that could be provide to assist in running a
> log linear regression would be greatly appreciated.
>
> I could not find anything in my manual about it.
>
> Thank you,
> cheryl
> *
> *   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:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/