s-news
[Top] [All Lists]

Estimation of Time Varying Parameters

To: <s-news@lists.biostat.wustl.edu>
Subject: Estimation of Time Varying Parameters
From: "Chen Sichong" <sichong.chen@postgrad.manchester.ac.uk>
Date: Sun, 6 Nov 2005 00:47:27 +0800
Thread-index: AcXiKJqzSgb0AwA/Q0qnzLmQn/3hdg==

Hi,

 

I am using nlminb to estimate a parameter which is time varying and autocorrelated by itself. Say, for example,

 

rho(t)=a0+alpha*rho(t-1)+beta*X(t-1)

 

Previously, rho is a fixed number, now I use the formula above to let it change with time. This, a0, alpha and beta is the parameters which I want to estimate. X(t) is a known time series vector.

 

What I program is to create a new vector rho, so I have to initial the first value of vector rho

 

rho=NULL

   rho[1]=0.2

      for(i in 2:(length(X)))

   {

       rho[i]=a0+alpha*rho[(i-1)]+beta*X[(i-1)]

   }

 

Then I put rho back to my ML function. As we know, to create a time varying vector using iteration is quite time consuming. Do you have some idea how to make an autocorrelated vector without iteration?

 

Many thanks,

Chen Sichong

 

PhD of Finance
Manchester Business School
University of Manchester
Oxford Road
Manchester
United Kingdom
M13 9PL

 

<Prev in Thread] Current Thread [Next in Thread>
  • Estimation of Time Varying Parameters, Chen Sichong <=