*
* Example of ARDL model starting page 251
*
open data inflation.dat
calendar(m) 1983:12
data(format=free,org=columns) 1983:12 2006:05 year month wage pcwage cpi infln
*
set y = 100*(log(cpi)-log(cpi{1}))
set x = 100*(log(wage)-log(wage{1}))
*
linreg y
# constant x{0 to 3} y{1 2}
@regcorrs(number=24)
*
* This is a bit more advanced. It uses the function %polydiv, which divides one
* polynomial by another. Since that will, in general, produce an infinite series,
* you have to indicate a truncation point - here, that's the 13.
*
* %EQNLAGPOLY(0,x) takes the last regression and extracts the lag polynomial for
* variable x. %EQNLAGPOLY(0,y) does the same thing for the variable y.
*
compute lagwgts=%polydiv(%eqnlagpoly(0,x),%eqnlagpoly(0,y),13)
*
* lagwgts is a VECTOR, which isn't the same thing as a SERIES. This copies the
* information out so we can use it in a GRAPH.
*
set dlweights 1 13 = lagwgts(t)
graph(footer="Figure 9.9 Distributed lag weights for ARDL model",$
 vlabel="Lag weight",hlabel="Lag",number=0)
# dlweights