12

Assume that we have a random variable $W = \max({X,Y})$ and that we would like to find the pdf of $W$. This is what I have done.

$$ F_W(w)= \mathbb{P}[ W\leq w]=\mathbb{P}[ \max({X,Y})\leq w]=\mathbb{P}[ X\leq w]\mathbb{P}[Y\leq w]= F_X(w)F_y(w) $$ then the pdf is $$f_W(w) = \frac{dF_W(w)}{dw}=\frac{d (F_X(w)F_y(w))}{dw}= f_x(w)F_y(w)+ f_y(w)F_x(w)$$

Is my reasoning correct?

What if one want to find the distribution of $W = \min({X,Y})$?

Franklin Pezzuti Dyer
  • 39,754
  • 9
  • 73
  • 166
Henry
  • 1,204

1 Answers1

10

$\color{green}{\checkmark}\quad$ Your reasoning for the density function of $W=\max(X,Y)$ is correct. Apply the same reasoning to find that of the minimum, with minor modification.

Vis: Let $Z=\min(X,Y)$ such that $X$ and $Y$ are independent and both continuous random variables, with density functions: $f_X$ and $f_Y$. (And cumulative distributions $F_X, F_Y$).

$\begin{align} f_Z(z) & = \frac{\mathrm d\;}{\mathrm d z} \Bbb P(\min(X,Y)\leq z) \\[1ex] & = \frac{\mathrm d\;}{\mathrm d z} (1 - \Bbb P(\min(X,Y)\gt z)) \\[1ex] & = -\frac{\mathrm d\;}{\mathrm d z}\Big( \Bbb P(X>z)\,\Bbb P(Y>z) \Big) \\[1ex] & = -\frac{\mathrm d\;}{\mathrm d z} \Big(\big(1-F_X(z)\big)\big(1-F_Y(z)\big)\Big) \\[1ex] & = f_X(z)\Big(1-F_Y(z)\Big) + \Big(1-F_X(z)\Big)f_Y(z) \end{align}$

Graham Kemp
  • 129,094
  • Hi @Graham could you give the intuition behind why subtract 1 - ℙ[min(,)>z] vs. just ℙ[max(,)≤] ? – Googme Oct 17 '20 at 06:36
  • The event of the minimum being less than a number is the event of at least one variable being so.$${\min(X,Y)< n}={X<n}\cup{Y<n}$$However, the event of the minimum being greater than a number is the event of both variables being greater than that number. $${\min(X,Y)> n}={X>n}\cap{Y>n}$$ The probability for the intersection of two independent events equals the product of the probabilities for each event. So working with this is useful. – Graham Kemp Oct 18 '20 at 06:26
  • For maximum, this is the other way around. The event of the maximum being less than a number is the event of both variable being so. Thus this is the best to work with..$${\max(X,Y)< n}={X<n}\cap{Y<n}$$ – Graham Kemp Oct 18 '20 at 06:30