Dear members!

I am running a log-log model (almost all variables are in logs) with panel data of T=28 and N=26.

Unfortunately, I am completely at lost about which model to use, I somehow always find something, which does not fit to a particular model.

I tested - to the best of my knowledge - all there is to test about the properties of my data and the following are the results:

- Hausman-test showed: fixed-effects
- time fixed-effects need to be included
- evidence of non-stationarity
- serial correlation
- cross-sectional dependence
- presence of heteroskedasticity

Does anyone have an idea which model they would employ?

Thank you so much for any propositions!

Regards,
Lara