Question

I would like to dynamically parameterize an array for a state-space model depending on how many states I choose.

I am doing this with a loop –

Q <- function(params,states) {
qmat <- matrix(0,statespace,statespace)
for (i in 1:statespace)
  qmat[i,i] <- statshockvar(params[(i-1)*5+1], params[(i-1)*5+2], 
                            params[(i-1)*5+3],states[i])          
qmat            

}

This function is called many times, as the point of the program is to optimize a paramter set. However, this function setup is slowing down the optimization phase very substantially because this function and a bunch of others like it keep getting called, and they keep redefining the arrays.

How can I define the arrays I need once, dynamically, with the relevant parameters as above, and then be able to call the matrix function with a new set of parameters for optimization?

Thanks!

Edit –

statespace is just an integer describing the number of states to use in the model, say 3/

statshockvar <- function(meanrev,longrun,sigma,sstate) {

longrun*sigma^2/(2*meanrev)*(1-exp(-longrun))^2+sigma^2/longrun*(exp(-longrun) - 
  exp(-2*longrun))*sstate

}

statshockvar – in this particular example is the discretized variance of a CIR model for the term structure

Edit 2 –

params looks like this – please note these are just arbitrary number

params = c(
0.3275,
0.07,
0.197,
0,
0.05,

0.01,
0.2,
0.3,
0,
0.05,

0.01,
0.1,
0.3,
0,
0.05)

states would be something like this –

states = c(0.07,0.07,0.07)

again these states are arbitrary.

Answer 1

Here’s a solution:

Q <- function(params, states) {
  diag(mapply(function(y, z) statshockvar(y[1], y[2], y[3], z), 
              lapply(seq(statespace), function(x) params[(x-1)*5 + 1:3]), 
              states))
}

Test with the example parameters:

Q(params, states)

            [,1]       [,2]        [,3]
[1,] 0.002465305 0.00000000 0.000000000
[2,] 0.000000000 0.03424762 0.000000000
[3,] 0.000000000 0.00000000 0.009499883