This is PART 2 of my questions about autocorrelation in software GRETL.
Hi, I´m student and I need to analyse the effect of monetary policy (represented by exchange rate, interest rate, money supply and indicator of systemic risk + dummy financial crisis) on output and inflation during crisis and after. I´ve chosen multiple regression analyses with time series.
Models with output are characterised with autocorrelation and most of them with heterogenity too. I tried to change functional forms in different combination, I used logarithm of dependent variable and finally first diferences for all model. After all of my effort autocorrelation is still there. I am trying to use Cochrane Orcutt method thanks to guide on youtube. Cochrane Orcutt in Gretl guide and I would like to know:
Is it necessary to use logarithm of all variables (both dependend and independent)?
Should I use CO method on all variables in my primary model? Or should I use model with different functional forms modified of variables with colinearity etc?
After using CO method, there is no more possible to use RESET test, and LM test or tests of heterogenity. Does that mean, that all this precondition (homogenity, well specified model etc) are fulfiled? And I have to take care only about normality and colinearity?
When I tried CO on my primary model (not modified, with all original variables) the only one significant variable was dummy crisis. Can I use dummy in CO, or is there any limitation of using it? Because if I don´t include dummy crisis into model, other variables are significant too. But in both cases there is not normality.
Thank you in advance for any response.
