This project is read-only.

Confidence Bands for Unbalanced Panel SVAR

Dec 3, 2014 at 4:30 AM
I haven't found anything that explicitly says that IRIS' toolbox cannot handle unbalanced data sets for panel VARs, so I assume that I've managed to get correct estimates (I do get an error message from the estimate command stating that "the following period(s) not fitted because of missing observations", but I'm taking that as more of a warning that it doesn't have data for ALL groups, rather than to say it's completely ignoring the data in those periods even though SOME groups may have observations).

My real issue is in trying to estimate confidence bands for the shock response functions. I'm trying to follow the bootstrap_VAR.m tutorial and adapt the "resample" command for a panel SVAR, but I'm running into problems because sending the full data range to resample causes it to have difficulty identifying the proper initial values for each country. So I've tried looping the resample command over all the countries using each country's unique date range of available data, but the resample command expects there to be more than one "group" when being using the SVAR from a panel.

Could someone give me a pointer on how to proceed? Or maybe resample just isn't compatible with an unbalanced panel SVAR?