Documentation: "more complex model" example

[gergelybat]

In the reference manual (KFAS.pdf) under "More complex model" the following model is defined: model <- SSModel(y ~ SSMregression(~ x1 + x2, Q = 0, R = matrix(c(1, 0), 2, 1)) + SSMarima(rep(0, 2), 0, Q = 0), H = 0) Later on the coefficient for x2 (beta2) is not estimated. In order to make it work, I made the following changes (but I'm new to KFAS, this might be not the right approach):

• in the update_function updated the corresponding parameter: model["a1", 3] <- pars[6]
• Made initial states of x2 and Intercept non-diffuse (perhaps Intercept could be removed from the model completely):

  model["P1inf", 3,3] <- 0
  model["P1inf", 1,1] <- 0

• called fitSSM() with 6 paratemers: fit <- fitSSM(model = model, inits = rep(0.1, 6), updatefn = update_function, method = "BFGS")

[helske]

beta2 is estimated similarly as the intercept and beta1, they are all part of the state vector, with variance of intercept and beta2 being constant. It is just I did not draw beta2 at the end as it is constant.

[gergelybat]

Thank you for the explanation. I really overcomplicated this.