Suppose I have the process $X = X(t)$ for $t \ge 0$ given by $X(t) = \sqrt{t}*Z \,\forall t \ge 0$ where $Z$ is normally distributed with $N(0,1)$.

Is this a Brownian motion?

Solution yields:

$$X(t)-X(s)=Z\sqrt{t} - Z\sqrt{s} \sim N\left( 0,(\sqrt{t}-\sqrt{s})^2)\right) = N\left(0,t-2*\sqrt{s\,t}+s\right)$$

and now we must compare with $X(t-s)$, etc.

However this is not my question, my question is how does $(Z\sqrt{t}-Z\sqrt{s})$ become $N(0,(\sqrt{t}-\sqrt{s})^2)$?

Why is it not $N(0,\sqrt{t}-\sqrt{s})$ only, without squaring?

So basically what does $Z(\sqrt{t}-\sqrt{s})$ mean intuitively and mathematically?

Would be grateful for any answer.

  • $\begingroup$ Also quite fundamental: does it even have independent increments? $\endgroup$ – Phil-ZXX Feb 16 '18 at 11:58
  • $\begingroup$ Your notations need to be clarified. Is $Z$ a single Gaussian random variable, in which case $X(t)=\sqrt{t/s} X(s)$ ? Or are the $Z$ independent draws of a Gaussian distribution (a new one for each $t$) ? In either case the $X$ is not a martingale and is not a Brownian motion. $\endgroup$ – Antoine Conze Feb 16 '18 at 15:41

If $Z \sim N\left( 0,1\right) $ then

\begin{align*} E \left(Z\sqrt{t} - Z\sqrt{s} \right) = 0\\ \end{align*}


\begin{align*} Var\left(Z\sqrt{t} - Z\sqrt{s} \right) = Var\left(Z(\sqrt{t} - \sqrt{s} )\right)=(\sqrt{t} - \sqrt{s})^2Var(Z)=(\sqrt{t} - \sqrt{s})^2 \end{align*}

Hence, since then sum of normals is normal

\begin{align*} Z\sqrt{t} - Z\sqrt{s} \sim N\left( 0,(\sqrt{t}-\sqrt{s})^2\right) \end{align*}

What this means is that you are defining a new process $Z$ which has a volatility $(\sqrt{t}-\sqrt{s})^2$

Hope this helps

| improve this answer | |

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.