So that the following is possible
y = dseries([0],'1990Q1','y');
e = dseries(randn(1000,1),'1990Q1','e');
from 1990Q2 to 2239Q4 do y(t) = .5*y(t-1) + e(t) - .5*e(t-1)*y(t-1) +.1*y(t-1)^2
Initially y has only one observation (which is mandatory because y
depends on its first lag), the routine extends the number of
observations so that y.dates(end)==2239Q4
Note that the exogenous variables are not adjusted, they must be
defined (leads/lags included) between the first and last dates of the
from-to-do syntax.