stsm (version 1.9)

methods-stsmFit: Methods to Extract Information from a Fitted stsm Model Object

Description

Common methods to print and display results for an object of class stsm or stsmFit returned by functions maxlik.fd and maxlik.fd.

Usage

"coef"(object, ...)
"print"(x, digits = max(3L, getOption("digits") - 3L),  vcov.type = c("hessian", "infomat", "OPG", "sandwich", "optimHessian"), ...)
"fitted"(object, std.rediduals = TRUE, version = c("KFKSDS", "stats"), ...)
"residuals"(object, standardised = FALSE, version = c("KFKSDS", "stats"), ...)
"plot"(x, ...)
"predict"(object, n.ahead = 1L, se.fit = TRUE,  version = c("KFKSDS", "stats"), ...)
"predict"(object, n.ahead = 1L, se.fit = TRUE,  version = c("KFKSDS", "stats"), ...)
"plot"(x, ...)
"tsSmooth"(object, version = c("KFKSDS", "stats"), ...)
"tsSmooth"(object, version = c("KFKSDS", "stats"), ...)
"plot"(x, ...)
"tsdiag"(object, gof.lag = 10L, ...)

Arguments

object

an object of class stsm or a stsmFit list.

x

a stsmFit list. For plot methods it is a list of class stsmComponents, stsmPredict or stsmSmooth.

digits

minimal number of significant digits, see print.default.

vcov.type

a character indicating the type of covariance matrix to be used to compute the standard errors of the parameter estimates.

version

a character indicating whether the Kalman filter and smoother functions from package KFKSDS or from the stats package should be used.

std.rediduals

logical. If TRUE residuals are standardised.

standardised

logical. If TRUE standardised are returned.

n.ahead

a numeric, number of observations ahead to perform prediction.

se.fit

logical. If TRUE standard errors of the predictions are returned.

gof.lag

numeric, number of lag autocorrelation coefficients to which apply the Box test.

type

A character. Type of information used to compute the covariance matrix of the parameters of the fitted model: information matrix, Hessian or a mixture of them.

...

additional arguments to be passed to the functions called in these methods.

Value

The following information is returned by these methods:

Details

These methods are based on those with the same name available for the output returned by StructTS in the stats package.

These methods are originally intended to provide summary information from a model fitted by maximum likelihood. Thus, the most natural input for them is a stsmFit list returned by maxlik.fd and maxlik.td. Nevertheless, as the information and the data required by these methods are available in the slots of a stsm object, they can also be applied directly on an object of class stsm. This can be useful, for example, when we know a set of parameter estimates that was obtained from a method other than maxlik.fd or maxlik.td. By simply updating the slot pars of the stsm object, the residuals and the filtered and smoothed components are readily available through these methods.

In most cases, the ellipsis, ..., is kept in the definitions of the methods just because it is part of the parent method. It has some functionality nonetheless. For the methods fitted.stsm and predict.stsm it can be used to pass argument P0cov to function char2numeric. It can also be used to pass graphical parameters to par in method plot.stsmComponents and plot.stsmSmooth or to plot in plot.stsmPredict and plot.stsmSmooth.

By default in method fitted, std.rediduals = TRUE so that it behaves as in previous versions of the package where this argument was not available. In method rediduals, by default standardised = FALSE because it is more convenient when used in package tsoutliers (this argument does not need to be explicitly defined and hence the method residuals is used in the same way both on arima models and for stsm).

See Also

confint.stsmFit, maxlik.fd, maxlik.td, vcov.stsmFit, stsm.

Examples

# fit the local level plus seasonal model to a 
# sample simulated series
data("llmseas")
m <- stsm.model(model = "llm+seas", y = llmseas)
res <- maxlik.fd.scoring(m = m, step = NULL, 
  information = "expected", control = list(maxit = 100, tol = 0.001))
print(res)

#diagnostic
tsdiag(res)

# display estimated components with 95% confidence bands
comps <- tsSmooth(res)
plot(comps)
title(main = "smoothed trend and seasonal components")

# plot predictions eight periods ahead
pred <- predict(res, n.ahead = 8, se.fit = TRUE)
plot(pred)