I'm conducting a study to check the impact on the wholesale electricity prices caused by the penetration of the renewable energies in the Portuguese electric system. I have hourly data from 2009 to 2018.
My dependent variable is the electricity prices and my explanatory variables are the electricity generation by source, electricity consumption, imports and exports.
Before running my regression I've already transform all my variables into logarithms, I've identified and corrected outliers and I have created dummy variables for day of the week, month, year and for each holiday (all accordingly to the existent literature).
After running my GARCH(1,1) model, I keep facing problems in the residual diagnostics, namely:
- - The p-values of the Correlogram Q-statistics all all zero (with 36 lags)
- The p-values of the Correlogram Square Residuals are all zero except for the first lag value (with 36 lags)
- In the ARCH test for Heteroskedasticity, the null hypothesis is rejected, but it is not rejected if I add a lag for my dependent variable in the regression [elec_price(-1)]
Can someone help solving this problem?
I thank you all in advance.