namedTrends: Specific trend models

In dajmcdon/myFKF: Fast Kalman Filtering and Smoothing

Description

This is simply a wrapper for trendModelArb which produces specific trend models discussed in Chapter 3 of Durbin and Koopman.

Usage

namedTrends(trendType, parmlist)

Arguments

trendType	The name of a specific trend type. Options are LocalLevel, LocalLinTrend, SmoothSlope, ConstantSlope, and IntRandWalk. Input should be a quoted string. This updates the parmlist to reflect the specific trend choice. Any other string simply applies trendModelArb to the parmlist.
parmlist	A list of parameters as generated by parmList.

Details

The named trend models have the following forms:

1. LocalLevel:

y_t = mu_t + more

mu_t+1 = mu_t + xi_t

2. LocalLinTrend:

y_t = mu_t + more

mu_t+1 = mu_t + nu_t + xi_t

nu_t+1 = nu_t + eta_t

3. SmoothSlope:

y_t = mu_t + more

mu_t+1 = mu_t + nu_t

nu_t+1 = nu_t + eta_t

4. ConstantSlope has trend = constSlope, gives linear trend

y_t = mu_t + more

mu_t+1 = constSlope + mu_t

5. IRW(r), r > 1,default is r=2

y_t = mu_t + more

mu_t+1 = 2mu_t - mu_t-1 + xi_t

Value

A list of matrices with components Z, Tt, Q, and R.

References

Durbin, James, and Koopman, Siem Jan. Time Series Analysis by State Space Methods (2nd Edition). Cambridge, GBR: OUP Oxford, 2012.

dajmcdon/myFKF documentation built on May 3, 2019, 5:16 p.m.