Background: I working on how GDP is affected by CPI, investment,consumption and commodity price. I am working with time-series

Problem: I tried to regress based on log values of all but this resulted in residuals being heteroskedastic and also R-squared was equal 1.0 which I had reservations about (something just didn't feel right). I then decided to change it by regressing log difference of all variables. The results were good with residuals being near perfect homoskedastic as well as stationary though not normally distributed but found some papers that argue normality may not be a problem.

My question is whether it's a good idea to pursue this further? I searched the internet for some answers but found nothing really so any opinions would really be helpful or any links to any research papers that might be useful

I am using Stata to do regression.

  • 1
    $\begingroup$ Are you using nominal GDP or real GDP? I think that one problem is that investment and consumption (in nominal terms) are dependent on CPI and they are components of GDP. This explains your strange $R^2$ value because GDP will fluctuate directly (one-to-one) with investment and consumption. I think you need to tweak your model significantly to not include the components of GDP in a regression for GDP $\endgroup$ – DornerA Apr 20 '16 at 15:29
  • $\begingroup$ @DornerA I will give it a go and get back to you. TY for reply. $\endgroup$ – Zhivago Apr 20 '16 at 15:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.