01-22-2016 04:44 PM
I try to find structural brake changes on variable like IPI or unemployment rate, firstly the BP test gives me too many breaks (like 10 in 30 years, impossible!) so I changed the EPS option in the code :
proc autoreg data=mem.ipiesptem; model e=e1 e2 / bp=(M=7, P=0, EPS=0.1, PRINTEST=ALL); run;
after it i obtained like 6 breaks (with the SupF(L+1) test it's still too many (but we can continue the analysis with this number) and the SupF test gives me more than 7 breaks.
1. Can I choose the number of break just with the SupF(L+1) test? (Because the other is always wrong.. and sequential test must be more robust than the simple SupF test)
2. Is the code adapted to my goal?
3. Is there a solution/modification of my code in order to limit the BP test to the most important breaks in this time series?
Many thanks for any help!!
01-25-2016 03:10 PM
This type of question is best answered by SAS Technical Support: http://support.sas.com/techsup/contact/. Best of luck to you, and thank you for using the SAS online communities.
Online Community Manager