ssm: Define state-space model in a glm-style call.
In ClausDethlefsen/sspir: State Space Models in R

Description Usage Arguments Value Author(s) See Also Examples

Use a glm-style formula and family arguments to setup a state space model.

ssm(formula, family = gaussian, data = list(), subset =
    NULL, fit = TRUE, phi = NULL, m0 = NULL, C0 = NULL,
                 Fmat = NULL, Gmat = NULL, Vmat = NULL, Wmat = NULL)
## S3 method for class 'ssm'
C0(ssm) 
## S3 method for class 'ssm'
m0(ssm) 
## S3 method for class 'ssm'
Fmat(ssm)
## S3 method for class 'ssm'
Gmat(ssm)
## S3 method for class 'ssm'
Vmat(ssm)
## S3 method for class 'ssm'
Wmat(ssm)
## S3 method for class 'ssm'
phi(ssm) 
## S3 replacement method for class 'ssm'
C0(ssm) <- value
## S3 replacement method for class 'ssm'
m0(ssm) <- value
## S3 replacement method for class 'ssm'
Fmat(ssm) <- value
## S3 replacement method for class 'ssm'
Gmat(ssm) <- value
## S3 replacement method for class 'ssm'
Vmat(ssm) <- value
## S3 replacement method for class 'ssm'
Wmat(ssm) <- value
## S3 replacement method for class 'ssm'
phi(ssm) <- value
getFit(ssm)

`formula`	a formula with univariate response on the lefthand side. The righthand side is a sum of terms and the special functions `sumseason`, `polytime`, `polytrig`, and `season` can be used. Terms can be marked by the `tvar`-function to create a term with time-varying coefficients. A special case is `tvar(1)` meaning a random walk.
`family`	a description of the error distribution and link function to be used in the model. This can be a character string naming a family function, a family function or the result of a call to a family function. (See `family` for details of family functions.)
`data`	an optional data frame containing the variables in the model. If not found in `data`, the variables are taken from `environment(formula)`, typically the environment from which `ssm` is called.
`subset`	an optional vector specifying a subset of observations to be used in the fitting process.
`ssm`	an object of class `ssm`.
`fit`	a logical. If `TRUE`, the model is fit using the `kfs` method. Otherwise, the model is returned as is.
`phi`	a vector of initial values of hyperparamters. Note that these have to be in the right order. Best advice is to leave this option to be `NULL` and then inspect the returned result using `phi(ssm)`.
`m0`	a vector with the initial state vector.
`C0`	a matrix with the variance matrix of the initial state.
`Fmat`	a function giving the regression matrix at a given timepoint.
`Gmat`	a function giving the evolution matrix at a given timepoint.
`Wmat`	a function giving the evolution variance matrix at a given timepoint.
`Vmat`	a function giving the observation variance matrix at a given timepoint.
`value`	an object to be assigned to the element of the state space model.

An object of class ssm with the following components

`ss`	an object of class `SS` describing the state space model. In addition, the `ss` object contains the components `family` and `ntotal` (for binomial case).

Claus Dethlefsen and Søren Lundbye-Christensen.

SS, extended

data(vandrivers)
vandrivers$y <- ts(vandrivers$y,start=1969,frequency=12)
vd.time <- time(vandrivers$y)
vd <- ssm( y ~ tvar(1) + seatbelt + sumseason(vd.time,12),
          family=poisson(link="log"),
          data=vandrivers,
          phi = c(1,0.0004),
          C0=diag(13)*100,
          fit=FALSE
          )
phi(vd)["(Intercept)"] <- exp(- 2*3.703307 )
C0(vd) <- diag(13)*1000
vd.res <- kfs(vd)

plot( vd.res$m[,1:3] )

attach(vandrivers)
plot(y,ylim=c(0,20))
lines(exp(vd.res$m[,1]+vd.res$m[,2]*seatbelt),lwd=2 )
detach(vandrivers)