Suppose you have a time series $\ Z_t$ that is used as a forecasting model for $\tau$ steps ahead from origin $T$. $\ Z_t$ is defined as:

$\ Z_t = 0.05 Z_T + 0.10 Z_{T-1} + 0.15 Z_{T-2} + 0.20 Z_{T-3} + 0.15 Z_{T-4} + 0.10 Z_{T-5} + 0.05 Z_{T-6 }$

How to calculate the $\ average\ age$ of this series observations in this forecasting model?

Is it right to assume the average age does not depend on the weights in each $\ Z_{T-i}$? In this case, the age of observation $\ Z_T$ is zero, since it's the observation at the origin $\ T$. The age of observation $\ Z_{T-1}$ is 1 since it happened one time-step ago. And so, ages will be respectively, from $\ T$ to $\ T-6$: 0, 1, 2, 3, 4, 5, and 6. Is this right?

And once the$\ average\ age$ is found, how to calculate a smoothing coefficent $\alpha$ that will yield a exponential smoothing model with the $\ average\ age$ equal to model $\ Z_t$?

If $\ Z_t$ were a exponential smoothing model, I'd know the value of $\alpha $; hence, I'd be able to calculate the average age of observations using equation: ($ \ 1$ - $\alpha $)/$\alpha $. However, it seems to me that $\ Z_t$ is a simple $\ AR$ model (moving average).


The average age of observations in a model is calculating by attributing age 0 for the current observation at time $t$, age 1 for the first past observation $t-1$, age 2 to the second past observation or the observation $z_{t-2}$, and so on.

In the end, if the model goes back to observation $z_{t-m}$ and it has one observation for all the intermediate periods between $z_t$ and $z_{t-m}$, the average age $AA$ will be:

$\large AA = \large \frac{\frac{(r_0 + r_m).m}{2} }{m+1}$

Knowing that, by the one of the criteria above, $AA(r_0)$ = 0, one can write:

$\large AA = \large \frac{\frac{(r_m).m}{2} }{m+1}$

where $r_m$ is the age of the last observation, and $m$ is the index of the last observation. So, if the last observation is $z_6$, $r_m$ = 6, and $m$ = 6.


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.