Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: bootstrap vector autoregression (var)?


From   Nick Cox <n.j.cox@durham.ac.uk>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: bootstrap vector autoregression (var)?
Date   Fri, 27 Jan 2012 15:57:28 +0000

The original -tsset- is not the issue here. Imagine the first bootstrap sample, which is taken with replacement. Then by chance some observations will be selected twice, or even more; and some not at all. So, -var- cannot operate as when it tries to operate it will find some times repeated, which is not allowed. So -var- will fail at the first hurdle. 

As you say, bootstrap commands in Stata specifically designed for time series are not in evidence. 

Nick 
n.j.cox@durham.ac.uk 

Saunders,Kyle

Stas et al,

Thanks for your reply.  I did indeed declare the -tsset- correctly ((tsset year, yearly delta(2)), and got results from the VAR and the Granger causality tests, so that's not the problem--unless you're saying that data need to have a delta of one (though I have 1972-2010, every two years).

I've done a pretty extensive literature search, and everything I find seems to have a bootstrap option for SVAR and VECM, but nothing for the straight VAR--and I'm trying to understand a) the difference between these procedures (I don't need the SVAR or the VECM), and b) why there isn't a bootstrap option for a straight up VAR--hence my question about this being a software issue, or if this is something about the VAR itself and how it differs from the others in the multivariate time series arsenal.

Also, I looked for user-written block bootstrapping modules, but they don't seem to be out there unless I am just missing them.





© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index