2
$\begingroup$

I am reading a paper where they prove Donsker's invariance principle for a sequence of dependent RV's. They do the following steps which I can't follow so well. I won't write out the precise definitions since I don't think they are needed, what I don't understand is the logic behind some steps.

Suppose $W_n(t)=n^{-\frac{1}{2}}\sum\limits_{k=1}^{[nt]}X_k$ where $t\in [0,1]$ is the sequence we want to show converges weakly to $W$ where $W$ is our Brownian motion.

This paper looks at a different sequence $M_n$ and shows that $M_n$ converges weakly to $W$. It then says that since for all $T>0$, $\left|\sup\limits_{t\in[0,T]}\left|W_n(t)-M_n(t)\right|\right|_{\mathcal{L}^2}\to 0$, it follows that $W_n\to W$ weakly.

This last step is the one I am struggling to understand. How does the assumption on $M_n$ ensure that $W_n\to W$? It appears to only show the convergence of FDD's (by Slutsky's theorem). Or does this also show tightness somehow? Or am I completely mistaken.

$\endgroup$
2
$\begingroup$

Please use following results in Billingsley's book: Convergence of Probability Measures, 2ed. p.124 Th3.1 p.27.
Theorem Suppose that $(X_n,Y_n)$ are random elements of $S\times S$. If $X_n\Rightarrow X$ and $\rho(X_n,Y_n)\Rightarrow 0$, then $Y_n\Rightarrow Y$.
Write $$ W_n(t)=M_n(t)+[W_n(t)-M_n(t)], $$ Then i) $M_n\stackrel{w}{\to}W$ in Skorokhod topology;

ii) $\mathsf{P}(\rho(W_n,M_n)>\varepsilon)\to0$, $\forall \varepsilon>0$, since $\mathsf{E}[\sup_t|W_n(t)-M_n(t)|^2]\to 0$ and $\mathsf{P}(\sup_t|W_n(t)-M_n(t)|>\varepsilon)\to0$, $\forall \varepsilon>0$.
Now by using Th.3.1 and from this two facts it is ready to get $W_n\stackrel{w}{\to}W$ in Skorokhod topology.

$\endgroup$

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.