1
$\begingroup$

So, I've got this exponential model: model-visual dr.

Which, when tested via Pagan- Breusch, got heteroskedasticity detected.


Breusch-Pagan / Cook-Weisberg test for heteroskedasticity Ho: Constant variance Variables: var2

     chi2(1)      =   760.75
     Prob > chi2  =   0.0000

And visualy:

...

My question, then, is: is there anything I can do to "save" my model?

$\endgroup$
1
  • 4
    $\begingroup$ Heteroscedasticity is not the chief problem: your exponential curve doesn't even remotely fit the data. $\endgroup$
    – whuber
    May 26, 2015 at 21:49

1 Answer 1

1
$\begingroup$

If you are regressing GDP time series then you should apply log-transformation to the GDP. Two popular ways of modeling GDP are:

$$\ln GDP_t=c + g t + \varepsilon_t$$ and $$\ln GDP_{t+1}-\ln GDP_t = g + \varepsilon_t$$

The first one is exponential growth, the second one geometric brownian motion. In any case you have to apply log, otherwise the heteroscedasticity will show up

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.