Skip to content

strucchange | weighted models

2 messages · Evan Cooch, Achim Zeileis

Evan Cooch
# 
Greetings -

Am hoping to use the strucchange package to look for structural breaks 
in some messy regression data. A series of preliminary analyses indicate 
that BLUE for these data will involve some weighting the data (estimates 
of a particular population parameter) by a function of the variance of 
the estimate (say, inverse of the variance). While I've gone through the 
docs for strucchange (which are excellent, btw), I don't see a simple 
(or obvious) way to apply some sort of 'weighting' to the regressions 
implemented in the package. Short of diving into source (which I could 
do, but I'm not sure how the various tests would be impacted by 
weighting of any sort), was wondering if anyone had dealt with this sort 
of issue - either with strucchange, or some other approach/package?

Thanks in advance...
Achim Zeileis
#
On Tue, 12 May 2009, f.query wrote:

            

      
      
      
    

        
Thanks!

            

      
      
      
    

        
I think there isn't in the "old" efp()/Fstats()/breakpoints() part, then 
there is no easy way. But int the "new" gefp() function you can use 
weights. If you want to do breakpoints estimation, I've got some modified 
code which is not included in the package...let me know if you need it.

hth,
Z