Characterizing invertibility of time series by natural filtrations

29 Views Asked by At

I'm studying linear time series and I'm thinking whether is possible to characterize an invertible moving average of order q (i.e.: MA(q) ) by means of filtrations generated by the two processes. More formally let

$$Y_t = \epsilon_t + \theta_1 \epsilon_{t-1}+\theta_2 \epsilon_{t-2} + ...+\theta_q \epsilon_{t-q}$$

where $(\epsilon_t)_{t \in N}$ and $(Y_t)_{t \in N}$ are both discrete time stochastic processes.

Can we state that:

$(Y_t)_{t \in N}$ invertible if and only if the filtration generated by the two processes at a fixed time $t$ are equal? More formally:

$$(Y_t)_{t \in N} \ invertible \iff F^Y_t=F^{\epsilon}_t \quad \forall t \in N$$

For $F_t^Y$ natural filtration of the process $(Y_t)_{t \in N}$ up to time $t$ (and same for $\epsilon$).

Sufficiency ($\Rightarrow$) seems to me pretty trivial consequence of basic measure theory. But I really have no clue about how to prove or disprove the necessity part.