Skip to content

Regression Time Series

5 messages · Ortega, Carlos (Carlos), Paul Johnson, Brian Ripley +1 more

Original
Ortega, Carlos (Carlos)
# 
Dear R-help,

I have been checking the different functions in the "ts" as well as
"tseries" libraries and I  could not find any reference to "regression time
series" or applying a model to a set of time series.

Please could you clarify if I missed any document or additional reference ?.

Thanks,
Carlos Ortega.



-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
Brian Ripley
#
On Wed, 4 Apr 2001, Ortega, Carlos (Carlos) wrote:

            

      
      
      
    

        
Look harder at arima0.

  
    

      
      
    

  


  


      

    

    
1 day later

    

      
      

        


  

        

      Paul Johnson
    

    

      
      #
    

  


  

      
I don't know what the original questioner intended, but...

In arima0 I find tools to fit AR, MA, and Integrated parts of a single
time series.  I was curious what you do if you want to do a multivariate
model. I see people talking about "intervention models" where some
variable y's been put through an ARIMA filter and then the effect of
independent variables x1,x2, etc, is assessed.  I think SAS calls these
things ARIMAX.  (ONline sas docs for version 8.1 can be seen here
http://lark.cc.ukans.edu/~sas81/sasdoc/sashtml/ets/chap7/sect1.htm)

I've not done these kind of models lately in SAS or R, but I've wondered
(as Carlos did) what R users do to estimate them.

        
Prof Brian Ripley wrote:

            

      
      
      
    

        
Paul E. Johnson                       email: pauljohn at ukans.edu
Dept. of Political Science            http://lark.cc.ukans.edu/~pauljohn
University of Kansas                  Office: (785) 864-9086
Lawrence, Kansas 66045                FAX: (785) 864-5700
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

  
  


  


      

    

    

      
      

        


  

        

      Brian Ripley
    

    

      
      #
    

  


  

      
On Fri, 6 Apr 2001, Paul E Johnson wrote:

        

            

      
      
      
    

        
Look even harder!  arima0 does fits ARIMAX models, using the xreg
parameter.

For an example, see ch13.R in the MASS/scripts collection.

ARIMAX models are not multivariate though: that's another subject
altogether.  For more complicated multiple models, bundle dse provides
a toolkit.

I know that the documentation is sparse, but this is an area due to be
replaced before too long.

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Paul Gilbert
    

    

      
      #
    

  


  

          

      
      
      
    

        
The terminology varies considerably and the number of letters gets quite confusing.
X usually means exogenous, which I do not think of as necessarily multivariate, but
some people do. A long time ago I reverted to using ARMA for just about everything
that includes the letters AR and MA. My dse package handles multiple endogenous
variables (y's) and multiple input or conditioning variables (which are often called
exogenous variables, but sometimes are not truly exogenous). Integrating models are
handled but you might need to consider what estimation technique to use. I have not
implemented any integration tests (volunteers?). DSE also handles state space models.
It does not implement time varying models or non-linear models, although it would
provide a valuable framework for anyone who would like to do that.

There should shortly be a new version of dse (dse_2001.4-1.tar.gz) on CRAN and it is
already available on my web site at <http://www.bank-banque-canada.ca/pgilbert>. The
User's Guide is also available at my web site. Relative to the version that has been
on CRAN the past few months this version fixes:

- format problems that left out many parts of the documentation
- maximum likelihood estimation (using optim but not extensively tested)
- and (I hope) problems that prevented it from working in Windows.

If anyone would like to compile and test the last item I would appreciate feedback.

Paul Gilbert

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._