All my variables are stationary, so cointigration can't be the problem. I have included one lag of the dependent variable and 5 explanatory variables. When I remove the explanatory variables, the R^2 doesnt change very much, only when I exclude the lag of the dependent variable, the R^2 is drastically reduced. I also checked the residuals and they dont seem to violate any assumptions.
Residual standard error: 0.005334 on 157 degrees of freedom Multiple R-squared: 0.968, Adjusted R-squared: 0.9664 F-statistic: 594 on 8 and 157 DF, p-value: < 2.2e-16
What could be the problem?
Can be because of over-fitting.
Please let me recommend our framework for forecast evaluation:
Conference Paper Data Formats and Visual Tools for Forecast Evaluation in Cyb...
If you have only 1 time series, you can use out-of-sample MAE to evaluate forecasting performance. R2 is not a good metric because it can be biased towards selecting over-fitted models.
Andrey Davydenko Thanks! I calculated the predicted R2 in order to check for overfitting. Doesn't seem to be a problem since there isnt a large discrepancy between predicted and adjusted R2.
Carl Kusche Are you using out-of-sample rolling-origin predictions to evaluate your model?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression