*
* Example 6.6.3 from pp 217-219
*
open data sbl.tsm
calendar(m) 1975
data(format=free,org=columns) 1975:1 1984:12 sbl
*
* There's a slight error in the description in the book, as the intervention
* variable should be 1 starting in t=98, not t=99.
*
* This creates a variable which is 1 when t is >= 1983:2 and 0 for time periods
* earlier than that.
*
set ft = t>=1983:2
*
linreg sbl
# constant ft
*
@bjident %resids
*
diff(sdiffs=1) sbl / dy
diff(sdiffs=1) ft  / gt
*
linreg dy
# gt
@bjident %resids
boxjenk(ma=12,reg,maxl,pmethod=simplex,piters=10,method=bfgs) dy
# gt
*
* An alternative way to specify this using RATS is to use the SDIFFS option to
* seasonally difference as part of the estimation. In that case, you put the
* standard regression function for the mean of the process, and include the
* "APPLY" option.
*
boxjenk(sdiffs=1,ma=12,maxl,regressors,apply,pmethod=simplex,method=bfgs) sbl
# ft