As I'm not able to comment (yet), my question follows the one raised by @mzuba here

I would like to use the DLM R package to model the local linear trend model, which unlike mzuba specified, has a stochastic trend.

The the model would be something like:

$y_t = x_t + ν_t, ν_t$~$N(0,V_t)$

$x_t = x_{t-1}+ z_{t-1} +ω_t,ω_t$~$N(0,W_t)$.

$z_t = z_{t-1}+ σ_t , σ_t$~$N(0,Σ_t)$.

I do have 2 questions regarding this model and the previous post:

  • In mzuba answer, the R code provided is dlmModPoly(order = 2, dV = exp(x[1]), dW = c(0,exp(x[2]))) while $W_t=\begin{pmatrix} x_2 & 0 \\ 0 & 0 \end{pmatrix}$. I would have naively expected the R code to have the two parameters inverted for $dW$. Could anyone tell me if I'm wrong?
  • In the case I'm right, I'd then expect my model to have $W_t=\begin{pmatrix} x_2 & 0 \\ 0 & x_3 \end{pmatrix}$ and the R code to be dlmModel <- dlmModPoly(order = 2, dV = exp(x[1]), dW = c(exp(x[2]),exp(x[3])))

Then the two output I'd get using dlmFilter(myData,dlmModel) would be a matrix with the first column being the filtered trend, and the second being the filtered slope.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.