MathOverflow will be down for maintenance for approximately 3 hours, starting Monday evening (06/24/2013) at approximately 9:00 PM Eastern time (UTC-4).

## Probability distributions: The maximum of a pair of iid draws, where the minimum is an order statistic of other minimums?

General question: What is the distribution for the maximum of 2 independent draws from cdf F(x), when we know that the minimum of those same two draws is the kth order statistic of the minimum of n pairs of independent draws from F(x)? Less technically, what is the distribution of the maximum associated with the kth greatest (of n) minima?

A specific example: Assume 8 independent draws from cdf F(x), which is defined over 0 to 1. Then, arbitrarily group the draws into 4 pairs. Compare the minimums of each pair. Label the maximum of these minimums as “a”. Label a’s pair (which is by definition > a) as “b”. Now, choose among the other three pairs arbitrarily, and label the two values in that pair as “c” and “d” (where c is the min of the pair and d is the max of the pair).

What are the distributions of b and d?

I know the distribution of a: F(a) = (1-(1-F(x))^2)^4 =Max of 4 draws of the Min of 2 draws of F(x).

I also know the distribution of c: F(c) = mixture of 1st , 2nd, and 3rd order statistics of 4 draws of Min of 2 draws of F(x). I get this by averaging the integrals (wrt x) for the pdfs that result from substituting (k=1, n=4), (k=2,n=4) and (k=3, n=4) into the following equation: (n!/((k - 1)!(n - k)!))(F(x)^(k - 1))*((1 - F(x))^(n - k))*F'(x)

I don’t know how to define F(b) or F(d)

And help would be greatly appreciated.

-
@Jennifer Did you look into extreme value theory books? – unknown (yahoo) Jul 18 2011 at 23:39
You could also look into asking this question at stats.stackexchange.com – David Roberts Jul 19 2011 at 4:17