Hello, fellow R users,
I recently need to estimate a dynamic linear model in the following form:
For the measurement equation:
Y_t = F_t * a_t + v_t
where Y_t is the observation. It is a 1 by q row vector for each t.
F_t is my forecasting variable. It is a 1 by p row vector.
a_t is my state variable. It is a p by q MATRIX of parameters with each column
of the matrix being regression coefficient of a random variable in Y_t. And v
is a multivariate normal noise.
The state equation is:
a_t = a_(t-1) + Omega_t
So a_t is a matrix random walk process. The distribution of Omega_t is a
matrix-normal distribution.
This model is very well discussed on page 579 of West and Harrison's book on
DLM. The key feature being that the state variable is a matrix. I have spent
quiet some time with package dlm and even bought the book written by the
authors from Amazon. But all the models disussed in the R book for dlm package
has the state variable being vectors. (For various reasons, I do not want to
stack up the columns of the matrix into a vector.) So I am wondering any body
can offer me some hints on how I could proceed? If anybody can show me an
example in R that would be fantastic.
Happy thanksgiving.
Wei
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
