*
* Example 10.3.3 from page 356
*
open data sunspots.dat
calendar 1770
data(format=free,org=columns) 1770:1 1869:1 sunspots
*
@bjautofit(pmax=5,qmax=5,demean) sunspots
*
* The ARMA(3,4) model has some local likelihood maxima, one of which BOXJENK
* finds rather than the (slightly) better set of coefficients shown in the text.
* We can get that one by starting out with conditional least squares estimation,
* then switching to ML.
*
boxjenk(ar=3,ma=4,demean) sunspots
boxjenk(ar=3,ma=4,demean,maxl,initial=%beta) sunspots
*
@bdindtests(number=20) %resids
*
set resids = %resids
*
* Fit GARCH model to the residuals. The default for GARCH includes an estimated
* intercept, which we want to omit here since we're working with residuals which
* should be close to mean zero.
*
garch(p=1,q=1,nomean) / resids