*
* Illustration from section 2.4.3
*
open data nile.dat
calendar 1871
data(format=free,org=columns,skips=1) 1871:1 1970:1 nile
*
* This is the same model as the previous example, but with type=smooth
* used to get smoothed rather than filtered estimates of the states and
* variances.
*
dlm(type=smooth,y=nile,a=1.0,c=1.0,sx0=1.e+7,sv=15099.0,sw=1469.1) / xstates vstates
set a = xstates(t)(1)
set p = vstates(t)(1,1)
set lower = a-sqrt(p)*%invnormal(.95)
set upper = a+sqrt(p)*%invnormal(.95)
*
* With the smoothed estimates, the first data point can be included in the
* graphs without distorting the picture.
*
spgraph(footer="Figure 2.2. Nile data and output of state smoothing recursion",hfields=2)
graph(header="Smoothed state and 90% confidence intervals") 4
# a 1871:1 *
# lower 1871:1 * 2
# upper 1871:1 * 2
# nile 1871:1 * 3
graph(header="Smoothed State Variance")
# p 1871:1 *
spgraph(done)