R help - Mar 2006 - Vector Autoregeressive Models: Adequation tests to perform

Hello,

 

I am currently testing a Vector AR of dim 3 over not a lot of data (135
* 3 observations) . To test the adequation of my vecot ar, I use the
Schwarz Bayesian Criterion and the classic modified Portmanteau test on
the residuals (it can be found for instance in
http://www.iue.it/PUB/ECO2004-8.pdf , page 15) -> the null hypothesis is
"the residuals process are a vectorila white noise process with
covariance matrix the one obtained from model calibration". I use the
mAr package.

 

My question is more statistical than purely r - related: If my order p
is, say, 12, what lag should I use in my portmanteau test? What is
usually  done in practice? 

And are there other tests that can be performed to judge the adequation
of the model?

 

Many thanks 

-------------- next part --------------


***********************************************************************
Bear Stearns is not responsible for any recommendation, solicitation, 
offer or agreement or any information about any transaction, customer 
account or account activity contained in this communication.
***********************************************************************

The "residual vector autocorrelation" paper you cite looks 
interesting, but I haven't used it, and unless you've received replies 
to the contrary that I haven't seen, I doubt if many people have much 
relevant experience.

	  Have you tried writing to one or more of the authors of that paper?

	  Alternatively, have you considered doing a Monte Carlo study?  If you 
already have code for the test you want to consider, it should not be 
too difficult to do.  One of the authors of the paper you cite might be 
happy to help you with designing a Monte Carlo study in exchange for 
co-authorship in a paper describing the results.

	  hope this helps,
	  spencer graves
p.s.  Do you have an answer to your March 9 post on modeling multiple 
realizations of an mAr process?  If no, are you familiar with Pinheiro 
and Bates (2000) Mixed-Effects Models in S and S-Plus (Springer) and the 
"nlme" and "lme4" packages?  Another alternative might be
the "AD Model
Builder" commercial software (www.otter-rsch.ca).  	

Zebouni, Stephane (Exchange) wrote:
> Hello,
> 
>  
> 
> I am currently testing a Vector AR of dim 3 over not a lot of data (135
> * 3 observations) . To test the adequation of my vecot ar, I use the
> Schwarz Bayesian Criterion and the classic modified Portmanteau test on
> the residuals (it can be found for instance in
> http://www.iue.it/PUB/ECO2004-8.pdf , page 15) -> the null hypothesis is
> "the residuals process are a vectorila white noise process with
> covariance matrix the one obtained from model calibration". I use the
> mAr package.
> 
>  
> 
> My question is more statistical than purely r - related: If my order p
> is, say, 12, what lag should I use in my portmanteau test? What is
> usually  done in practice? 
> 
> And are there other tests that can be performed to judge the adequation
> of the model?
> 
>  
> 
> Many thanks 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> 
> 
> ***********************************************************************
> Bear Stearns is not responsible for any recommendation, solicitation, 
> offer or agreement or any information about any transaction, customer 
> account or account activity contained in this communication.
> ***********************************************************************
> 
> 
> 
> ------------------------------------------------------------------------
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html