Exponential smoothing with damped trend

[Marnix Zoutenbier]

Dear all,

I would like to perform exponential smoothing with a damped trend on my data. I can not find the proper Stata-command(s) to do the task. Do I overlook something?

Thank you for your consideration.

Best regards,

Marnix Zoutenbier

[Nick Cox]

help tssmooth

[Marnix Zoutenbier]

Dear Nick,

Thank you for your help. As a result, I reread the help and also the TS-manual on tssmooth. However, I was not able to find what I was looking for.

Most closely to what I am looking for comes the Holt-Winters method which is basically described on page 671 with the formulas
x_hat(t+tau) = a(t) + b(t)*tau
a(t) = alpha*x(t) + (1-alpha)(a(t-1) + b(t-1))
b(t) = beta*(a(t) - a(t-1)) + (1-beta)*b(t-1)

What I am missing in these formulas is an 'extra' damping parameter, say `fie', in the following way
x_hat(t+tau) = a(t) + SUM{i=1...tau} b(t)*(fie)^i
a(t) = alpha*x(t) + (1-alpha)(a(t-1) + fie*b(t-1))
b(t) = beta*(a(t) - a(t-1)) + (1-beta)*fie*b(t-1)

Did I overlook this extension of the Holt-Winters method in the manual or does Stata indeed not support this model?

Thank you for your consideration,

Best regards,

Marnix Zoutenbier

[Nick Cox]

Sorry, I don't know either way. Time series models that look different often turn out to be identical, so there may be another method.

[Marnix Zoutenbier]

Dear Nick,

Thank you for your quick response. You are right that terminology in time-series is often different although the models are the same. I am going to dig a little bit deeper.

Best regards,

Marnix