1
$\begingroup$

I am having a small doubt regarding maximum of random variables. I have $$Z= \max\{ X_1, X_2,\dots X_p, \dots X_N\}$$ where all $X_i$ are independent, identically distributed. Now, If for sure, I know that $$\{X_1,\dots,X_p\}>X_{p+1}>\dots>X_N$$ the can i say that the CDF of $Z$,

$$\begin{align*}P(Z<z)&= P\left(\max\{ X_1, X_2,\dots X_p, \dots X_N\}<z\right)\\&=P\left(\max\{ X_1, X_2,\dots X_p\}<z\right)\\&=\left[P(X_1<z)\right]^{N-p-1}\end{align*}$$

How to find CDF of $Z$ in this case ? Is there any theorem that reduce the size of total variables for certain condition ?

$\endgroup$
1
$\begingroup$

If you can say for sure that the event $$E:=\left\lbrace\{X_1,\dots,X_p\}>X_{p+1}>\dots>X_N\right\rbrace$$ occurs, then you can express this as $$P(E)=P\left(\{X_1,\dots,X_p\}>X_{p+1}>\dots>X_N\right)=1$$ Therefore the distribution of $Z$ given this information can be expressed as $$\begin{align*}F_{Z\mid E}(z|E)&=\frac{P(Z<z, E)}{P(E)}\overset{P(E)=1}=\\&=P\left(\max\{X_1, X_2, \ldots X_N\}<z \, \text { and } \, \left\lbrace\{X_1,\dots,X_p\}>X_{p+1}>\dots>X_N\right\rbrace\right)\\&= P\left(\max\{X_1, X_2, \ldots X_p\}<z\right)=\ldots\\&=\left[F_{X_1}(z)\right]^{N-p-1}\end{align*}$$ as you have it. So, concerning your question, the notion of the conditional distribution is what you need in order to express your correct conclusion.

$\endgroup$

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.