# Convergence of a process

this may be viewed as a duplicate of this post.

However i have put in much effort in the shared link and donated it with reputation, to check the proof considered there.

Here however i want to argue about another alernative proof, which may be usefull.

Lets say for a zero mean martingale $(X_{t})_{t\geq 0}$ and $Var[X_t]=1$ we have $$\frac{X_t}{\sqrt{t}}\xrightarrow{d} \mathcal{N}(0,1)$$ (this is the case for mild assumptions of a process with stationary independent icrements). Lets say, if for a continuous increasing positive process $U_{t}$ and positive continuous increasing function $n(t)$ with $U_t,n(t)\rightarrow \infty$ as $t\rightarrow \infty$ it holds $$U_{t}/n(t)\xrightarrow{P} Y\quad a.s.$$ for a strict positive and finite random variable.

I want to show that, $$\frac{X_{U_t}}{\sqrt{U_t}}\xrightarrow{d}\mathcal{N}(0,1) \tag1$$ holds.

Is it sufficient to show (1), that for every constant $0<\eta<\infty$ where now $$U_{t}/n(t)\rightarrow \eta \quad a.s.$$ it holds, that $$\frac{X_{U_t}}{\sqrt{U_t}}\xrightarrow{d} \mathcal{N}(0,1)\,.?$$ IF NOT here is the alternative: According to @Did notice we have that $$g \left(\frac{X_{n(t)}}{\sqrt{n(t)}},\frac{U_t}{n(t)}\right)\xrightarrow{d} g (Z,\theta)$$ for any continuous function $g$ where $Z\sim\mathcal{N}(0,1)$ and $Z$ and $\theta$ are independent. Lets define $X_t \circ{} m:=X_{t\cdot m}$ we change the time. Then with $g(x,y)=\sqrt{y}^{-1/2}x\circ y$ we have $$\frac{\sqrt{n(t)}}{U_t}\frac{X_{U_t}}{n(t)}\rightarrow \sqrt{\theta}^{-1/2} Z\circ \theta$$ It would be nice that the last expression is $\mathcal{N}(0,1)$ distributed. Since $\theta$ and $Z$ are independent we can think of a time scaled Brownian motion $Z=W_{1}$ where $W_{t}$ is a brownian motion at time t. Where for $\sqrt{\theta}^{-1}W_{\theta} \sim\mathcal{N}(0,\sigma^{2})$. Or could we use a better function $g()$?

• With no independence hypothesis between $(X_t)$ and $(U_t)$, the result is doubtful. (But isn't this a duplicate of your other recent question?) – Did Jun 2 '16 at 8:27
• @Did the result is proven for given assumptions see. Basawa Statistical Inference for stochastic Processes Theorem 4.2 Appendix 1 page 395. You are right this is a specific question according to my main question here math.stackexchange.com/questions/1806716/… . However i thought maybe the most readers are scarred by the massive text, so i decided to ask this unclear part seperatly? Or isn't that allowed? – ziT Jun 2 '16 at 8:33
• @Did One can show that the Limits $(X_{t},U_t/n(t)\xrightarrow{d}(Z,Y)$ with $Z=\mathcal{N}(0,1)$ $Z$ and $Y$ are independent. Does this help? – ziT Jun 2 '16 at 10:12