Hi! I am using Stata 16. My data is daily panel data of 380 days and 185 countries. I have reviewed descent literature on information criteria. I understand them, but could not find a reasonable explanation for the lag selection for a large T and Large N. I am looking for a systematic solution to this. I know some packages can estimate AIC/SIC for model fitting purpose but not the optimal lag. Any specific guidance is much appreciated.
Shahnawaz Mohammad Rafi, it seems that the use of automatic procedures and rules of thumb has hidden the true purpose of an "optimal" lag: that is, whiten the residuals. The optimal lag is the one that removes correlation from residuals. Comparing AIC and BIC has sense only for models with uncorrelated residuals.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication