Hello sir/ma'am,
My study will employ seven stock indices (Daily) for a period of 10 years. The optimal lag length is 3 based on AIC. It means that there are 21 predictors. My question is is there maximum variables in VAR analysis? Because 21 predictors are too many and it may have effect on the degree of freedom.
Thank you
Hello
The number of variables depends on the method of study. If a regression model is built, some variables may be statistically insignificant. In any case, tests for unnecessary and missed variables can help
Try to define the number of lags, and begin to check whether the model with a certain number of lags will pass the diagnostics check or no!
Bear in mind that in VAR, you should order first the most exogenous variable.
Best, William Sucuahi
Adelajda Matuka all variables are considered endogenous in my study.
You have daily data on 10 years. The main problem in econometric models is the freedom degree. This is not a problem in your case. Besides VAR models are build to manipulate several variables taking also into account their lags. You have to try several lag orders and compare your results using a diagnostic on the lags coefficients significance, the likelihood and the information criterion indices. The choice of the lag order (in your case 3). I'm quite sure that if you use different criterions (AIC,BIC,LL, SBIC.........) you will get different lag orders as optimal choice you have to try them all and then decide the best choice.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis