Page 1 of 1

Anti-logs

Posted: Sat Jun 29, 2013 8:03 pm
by axiom
I want to anti-log an equation (it is in a double-log specification) so that I can compare it with its linear specification equivalent equation (I want to compare goodness of fit for alternative specifications of dependent variable).

Is there a simple way of doing this? I tried using an example that my former professor gave, but it doesn't appear to be working in EViews.

Re: Anti-logs

Posted: Sat Jun 29, 2013 8:12 pm
by EViews Gareth
You mean the @exp function?

Re: Anti-logs

Posted: Sun Jun 30, 2013 6:17 pm
by axiom
Yes, it has to do with the @exp function. Let me be more specific.

I used the following command, ls carsreal c persdispinc/cpi cpiautos/cpi intrate unemprate, to get the output below:
linear.png
linear.png (16.54 KiB) Viewed 4578 times
[img]

Note the following from the output above:
Adjusted R-squared 0.930680
S.E. of regression 2041.876
Sum squared resid 4.29E+08

Now, I also used the following command, ls log(carsreal) c log(persdispinc/cpi) log(cpiautos/cpi) log(intrate) log(unemprate), to get the log specification of my above equation:
log.png
log.png (16.93 KiB) Viewed 4578 times


Note the following from the output above:
Adjusted R-squared 0.892040
S.E. of regression 0.097696
Sum squared resid 0.983076

What I need to do is compare my log-specified equation with my linear equation. My professor gave a long procedure, which involves using the @exp function, but I am running into errors. What is a way I can compare their goodness-of-fit?

Re: Anti-logs

Posted: Tue Jul 02, 2013 10:46 am
by EViews Glenn
Can't you just use the forecast command to obtain the fits in levels and then go from there. The forecast will give you the RMSE, Theil and MAPE statistics as well.