I have a such example in my text book

Let Y and Z be two IID random variables each with CDF, $F(\cdot)$. Let $X_1:=min(Y,Z)$ and $X_2:=max(Y,Z)$ with marginals $F_1(\cdot)$ and $F_2(\cdot)$, respectively. We then have \begin{align} P\left(X_{1} \leq x_{1}, X_{2} \leq x_{2}\right)=2 F\left(\min \left\{x_{1}, x_{2}\right\}\right) F\left(x_{2}\right)-F\left(\min \left\{x_{1}, x_{2}\right\}\right)^{2}. \end{align} We can derive this by considering separately the two cases $\text { (i) } x_{2} \leq x_{1} \text { and (ii) } x_{2}>x_{1}$

We would like to compute the copula, $C\left(u_{1}, u_{2}\right) \text {, of }\left(X_{1}, X_{2}\right)$. Towards this end we rst note the two marginals satisfy

\begin{align} \begin{aligned} F_{1}(x) &=2 F(x)-F(x)^{2} \\ F_{2}(x) &=F(x)^{2} \end{aligned} \end{align} But Sklar's Theorem states that$C(\cdot,\cdot)$ satisfies $C\left(F_{1}\left(x_{1}\right), F_{2}\left(x_{2}\right)\right)=F\left(x_{1}, x_{2}\right)$ so if we connect the pieces we will obtain $C\left(u_{1}, u_{2}\right)=2 \min \left\{1-\sqrt{1-u_{1}}, \sqrt{u_{2}}\right\} \sqrt{u_{2}}-\min \left\{1-\sqrt{1-u_{1}}, \sqrt{u_{2}}\right\}^{2}$

My question is how does the Joint CDF $P\left(X_{1} \leq x_{1}, X_{2} \leq x_{2}\right)$ is equal to $2 F\left(\min \left\{x_{1}, x_{2}\right\}\right) F\left(x_{2}\right)-F\left(\min \left\{x_{1}, x_{2}\right\}\right)^{2}$ by using $\text { (i) } x_{2} \leq x_{1} \text { and (ii) } x_{2}>x_{1}$


When $x_2 \leq x_1$ we have $X_1 \leq x_1$ and $X_2 \leq x_2$ if and only if $Y \leq x_2$ and $Z \leq x_2$. [Note that $Y \leq x_2$ implies $Y \leq x_1$ and so $X_1 \leq x_1$ also]. Hence, $P(X_1 \leq x_1$ and $X_2 \leq x_2)=P(Y \leq x_2,Z \leq x_2)=F(x_2)^{2}$. Also, $2 F\left(\min \left\{x_{1}, x_{2}\right\}\right) F\left(x_{2}\right)-F\left(\min \left\{x_{1}, x_{2}\right\}\right)^{2}=2F(x_2)F(x_2)-(F(x_2))^{2}=(F(x_2))^{2}$.

Now let $x_1 <x_2$. Then $P(X_1 \leq x_1, X_2\leq x_2)$ can be split as $P(Y\leq x_1,Z \leq x_2)+P(Y>x_1, Z\leq x_1, Y\leq x_2)$. So we get $F(x_1)F(x_2)+(F(x_2)-F(x_1)) (F(x_1))=2F(x_1)F(x_2)-(F(x_1))^{2}$.

  • $\begingroup$ Thank you for your response. I have now tried to do the proof for $x_1 < x_2$. I find that evaluating the formula yields $$P(X_1\leq x_1, X_2\leq x_2)=2F(x_1)F(x_2)-F(x_1)^2$$ I can reason that $x_2>x_1$ allow for the events $(Y\leq x_1) \cap (Z\leq x_2)$ and $(Y\leq x_2) \cap (Z\leq x_1)$ that yields the $2F(x_1)F(x_2)$, but I cannot reason how to get the $-F(x_1)^2$. $\endgroup$
    – Qi Yao
    Sep 14 at 13:29
  • $\begingroup$ @QiYao I have added a proof for this case . $\endgroup$ Sep 14 at 23:19
  • $\begingroup$ Thanks for your answer, an last question, why we need $Y>x_1$ but not $Z>x_1$?? thanks alot $\endgroup$
    – Qi Yao
    Sep 16 at 8:18
  • $\begingroup$ @QiYao I am splitting $(X_1 \leq x_1, X_2\leq x_2)$ into $(Y \leq x_1,X_1 \leq x_1, X_2\leq x_2)$ and $(Y > x_1,X_1 \leq x_1, X_2\leq x_2)$. When $Y >x_1$ we need $Z \leq x_1$ so that $X_1 \leq x_1$. Note that $Z \leq x_2$ follows from $Z \leq x_1$ since $x_1<x_2$. $\endgroup$ Sep 16 at 8:38
  • $\begingroup$ thank you so much! $\endgroup$
    – Qi Yao
    Sep 16 at 8:52

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.