I am fitting dynamic linear models with the r package dlm. The text book associated the package, Dynamic Linear Models With R by Petris, Petrone and Campagnoli (2009) provides good and in detail explanation of to specify the models in R. When it comes to modelling the evolution variance (Wt in their notation) with variance discounting they provide a function on the book web page http://definetti.uark.edu/dlm/.
However, if I have a model which contains two components (trend/mean + seasonal for example) I would probably like to use different discounting factors for each component. Example:
library(dlm)
mod <- dlmModPoly(1) + dlmModSeas(4)
Here I have a local level model + seasonal component with quarterly period. Has anyone any idea how to specify different discount factors for each component in the Kalman Filter?
