the double exponential model is so simple:

level: $s_t = \alpha x_t + (1-\alpha)(s_{t-1}+b_{t-1})$

trend: $b_t = \beta (s_t - s_{t-1}) + (1-\beta)b_{t-1}$

both intuitively weigh the new information vs the existing info.

2/3 of the triple exponential model (additive seasonality for simplicity) are also intuitive and simple:

seasonality: $c_t = \gamma(x_t - s_{t-1} - b_{t-1}) + (1-\gamma)c_{t-L}$

trend: $b_t = \beta (s_t - s_{t-1}) + (1-\beta)b_{t-1}$

but the level equation is so weird!

level: $s_t = \alpha (x_t-c_{t-L}) + (1-\alpha)(s_{t-1}+b_{t-1})$

why is the new information part taking into consideration only the seasonal component while the old information part takes into consideration only the trend component?


it's not weird, it's just a weighted average between the seasonality adjusted observation: $x_t - c_{t-L}$ and the non-seasonal forecast for time t: $s_{t-1} + b_{t-1}$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.