For a given time series (x_{1}, x_{2}, ldots), the (N)-period arithmetic Moving Average (MA), (A(n, N)), is

Question:

For a given time series \(x_{1}, x_{2}, \ldots\), the \(N\)-period arithmetic Moving Average (MA), \(A(n, N)\), is defined as

\[A(n, N)= \begin{cases}\left(x_{1}+\ldots+x_{n}ight) / n & \text { if } n

Similarly, given \(0<\alpha<1\), the Exponential Moving Average (EMA), \(E(n, \alpha)\), is defined as

\[E(n, \alpha)= \begin{cases}x_{1} & n=1 \\ \alpha x_{n}+(1-\alpha) E(n-1, \alpha) & n>1\end{cases}\]

When computing \(A(n, N), E(n, \alpha)\), the age of a data point \(x_{n-i}\) is \(i\), and the Average Age is the weighted average age.

(a) Compute the average age of an \(\mathrm{N}\)-period arithmetic MA.

(b) Prove the following

\[E(n, \alpha)=\sum_{i=0}^{n-2} \alpha(1-\alpha)^{i} x_{n-i}+(1-\alpha)^{n-1} x_{1}\]
and verify the following asymptotic approximation for large \(n\)
\[E(n, \alpha) \approx \sum_{i=0}^{\infty} \alpha(1-\alpha)^{i} x_{n-i}\]

(c) Using the approximation from part (b), verify that the weights \(\alpha(1-\alpha)^{i}\) add up to 1 , and compute the asymptotic Average Age of EMA \[\sum_{i=0}^{\infty} i \alpha(1-\alpha)^{i}\]

(d) An EMA with an asymptotic average age equal to an \(\mathrm{N}\)-period arithmetic MA is called an \(\mathrm{N}\)-period EMA. What is the \(\alpha\) of an \(\mathrm{N}\)-period EMA?

(e) A crypto exchange defines its settlement price as the 30-second EMA of the Bitcoin price. What weight is assigned to the Bitcoin price of 15 seconds ago?

Fantastic news! We've Found the answer you've been seeking!

Step by Step Answer:

Question Posted: