i have problem with ETS smoothing, the output series do not chow level trend and seas just forecast when i chose different model...etc. my best model with auto specification chose (M,M,M) with AIC it chow just forecast series, it's the same problem.
have you an idea.
2nd question:
A/
in user guide II page 618 state space is:
@signal log(passenger) = c(1) + sv1 + c(4)*sv2
@state sv1 = c(2)*sv1(-1) + c(3)*sv2(-1) + [var=exp(c(5))]
@state sv2 = sv1(-1)
but when i repeat this example step by step by a new object and auto specification, i put log(passenger) in dependent variables and i put 'c' in regressor with fix coefficients for constant c(1) next i chose ARMA(2,1) next ok it chow this:
@signal log(passenger) = c(1) + sv1 + c(2)*sv2
@state sv1 = c(4)*sv1(-1) + c(5)*sv2(-1) + [var = exp(c(3))]
@state sv2 = sv1(-1)
the coefficients c(2) c(3) c(4) c(5) are not in order like page 618, how to write the coefficient in order when i use auto specification?
B/
it is difficult to write SARIMA(p,d,q)*(P,D,Q)*s in state space model and it is boring
when p,q and P,Q take different value, can you add 'SAR' 'SMA' 'D' 's' in auto specification.best regards