Discontinuous process with same marginals as Brownian Motion

64 Views Asked by At

I understand that there are many discontinuous modifications of Brownian Motion, one of which seen in my course was:

$$X_t = \begin{cases} B_t & t\neq u \\ 0 & t=u \\ \end{cases} $$

where $(B(t):t\in[0,\infty))$ is BM and $u$ is $\text{Unif}([0,1])$. However, I don't understand the claim that $(X(t))$ has the same finite dimensional marginals as $(B(t))$. Recall that same finite-dimensional marginals means that with probability 1, for any finite sets $\{t_j\}_1^k\subseteq T$, the collection of marginals of both stochastic processes equate in law with each other. Any help is appreciated.

Edit: I'm adding "with probability 1" to the beginning the definition of same finite-dimensional marginals for congruence with the answer below. Also, in the same light as the answer below, it's clear that if the BM was $(B(t): t\in [0,n] \cap \mathbb{N})$ for some $n$, then equality in f.d. marginals would no longer hold.

1

There are 1 best solutions below

1
On

The reason is that for any finite set on which you're taking marginals, the uniform random variable will almost surely never be on one of those points - this is true for a simpler example too:

For $t \in [0,1]$, take $X_t := 0$, a uniform $[0,1]$ random variable $U$, and $Y_t = \chi_\{U\}$ to be 1 at the value chosen by $U$, 0 otherwise. Then for any finite set $\{t_i\}$, we have $$ \mathbb P(Y_{t_i} = 0 \textrm{ for all }i) = 1 $$ so that $\mathbb P(X_{t_i} = Y_{t_i} \textrm{ for all }i) = 1$, but that $\mathbb P(X_t = Y_t \textrm{ for all }t) = 0$ since $X_U = 0 \not = Y_U = 1$ (there's always exactly one $t$ for which the two processes differ)

Your example just has more complicated (but still identical) definitions of processes away from the uniformly selected point.