3
$\begingroup$

I'm currently reading some time series lecture notes. It says that:

Weakly stationary (or wide-sense stationary) processes are said to be $I(0)$ (integrated of order $0$).

Let's call the above definition A.

On the other hand the Wikipedia page on order of integration says

A time series is integrated of order $0$ if it admits a moving average representation with $\sum _{k=0} ^\infty |b_k|^2 < \infty $, where $b$ is the vector of moving average weights. This is a necessary but not sufficient condition for a processes to be stationary.

Let's call the above definition B.

My question is: I am unsure if if definitions A and B for $I(0)$ are equivalent. Can someone please point me to a proof or a counter example? Many thanks!

$\endgroup$
  • $\begingroup$ Perhaps adding in exactly what "integrated of order $0$" means to you as well as an exact reference to where "wiki" gives an alternative definition of $I(0)$ will be helpful.... $\endgroup$ – Dilip Sarwate Jun 15 '16 at 12:39
  • $\begingroup$ @DilipSarwate, yes you are certainly correct that my post was badly organized. It is now edited. Thank you. $\endgroup$ – pippy Jun 15 '16 at 13:05
1
$\begingroup$

The definition I see there (and in the book referenced therein) is $(1-L)^dx_t=z_t$, and a process is of order 0 if $(1-L)^0x_t$ is stationary (when used without prefix it means strict stationarity). But when $d=0$ we simply get $z_t=x_t$, so a process is of order 0 if and only if it is stationary.

From this definition, $I(0)\Rightarrow$ weakly stationary but not vice versa, i.e. $I(0)$ is a sufficient condition for stationarity, but weak stationarity is only a necessary condition for $I(0)$.

$\endgroup$
  • 1
    $\begingroup$ To amplify what yoki wrote, I(0) means stationary. Weakly stationary means that the first two moments of a process (mean and variance) are time independent, where as stationary means that ALL moments are time invariant. $\endgroup$ – aginensky Jun 15 '16 at 13:46
  • $\begingroup$ @aginensky, I(0) and stationary are not equivalent. See the beginning of chapter 3 (e.g. p. 35) in Johansen's book "Likelihood-based inference in cointegrated vector autoregressive models". $\endgroup$ – Richard Hardy Jul 10 '16 at 19:53
  • $\begingroup$ @Hardy- embarassing mis- statement on my part. I merely meant to point out the second sentence on. $\endgroup$ – aginensky Jul 12 '16 at 3:38

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.