That is because the Kalman filter is only an estimation method. You will have to
specify the
matrices A and C in:
y_t = A_t * x_t + noise_t
x_t = C *x_t-1 + noise_t
where you plug the explanatory variables in A_t. The Kalman filter will then
give you estimates
of the state vector x_t...which includes your time varying coefficients. You
will also need to
add an optimisation routine on top of the Kalman filter to estimate any unknown
parameters.
Not difficult...requires a bit of programming.
Alain
---------------------
Dr Alain Zuur
Highland Statistics Ltd.
www.highstat.com
www.brodgar.com
Message: 3
Date: Wed, 7 Apr 2004 11:52:36 +0100
From: Arnaud_Amsellem@ssga.com
Subject: [R] Time Varying Coefficients
To: r-help@stat.math.ethz.ch
Message-ID:
<OF82C114A6.D20745CD-ON80256E6F.003B34AF-80256E6F.003BBFCE@statestr.com>
Content-Type: text/plain; charset=us-ascii
I'd like to estimate time varying coefficients in a linear regression using
a Kalman filter.
Even if the Kalman Filter seems to be available in some packages I can't
figure out how to use it to estimate the coefficients.
Is there anyway to do that in R?
Any help appreciated
Thanks