I am running GMM system with HAC weights, which is probably misspecified (NKDSGE) and therefore convergence behaves badly. However, despite misspecification, the model gives reasonable forecasts and that is what I am after.
The problem is, from time to time something change, and the results of exactly the same estimations are suddenly different. I first noticed it after the program crashed and when I re-run it, but later again with some change in estimation options (change of Kernel, when switched back to default, somewhat different results). I am not able to replicate my result therefore!
I use "param" to ensure identical starting values for iterative estimation method. Also tried direct edit of c.
Seems like if there were some other hidden values that affect the procedure of iterative GMM estimation, that sometimes change with some reinicialization of the program (and sometimes not).
I use Eviews 8, however not up do date, as this is not my copy.
How can I assure replicable results?
Changing estimated values
Moderators: EViews Gareth, EViews Moderator
Who is online
Users browsing this forum: No registered users and 21 guests