Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $T_k \sim E(q_k)$ be a countable family of independent exponentially distributed random variables. In a book by Norris (, at pg. 72, I found the following identity:

$ \ \ \mathbb{P}(T_k\geq t\text{ and } T_j>T_k\text{ for all } j\neq k) = \int_t^\infty q_k e^{-q_ks} \mathbb{P}(T_j > s \text{ for all } j\neq k) ds, $

which I'm not able to prove rigorously. Any hint is much appreciated.

share|cite|improve this question
Hint: For one approach, what is the distribution of the minimum of a collection of independent exponential random variables? – cardinal Feb 24 '12 at 16:19
up vote 0 down vote accepted

Let us rewrite the problem to emphasize the useful assumptions:

Let $X$ and $(Y_i)_i$ denote some random variables. Assume that the family $(Y_i)_i$ is countable and independent on $X$ and that the PDF of $X$ is $f$. Then, for every $z$, $$ \mathrm P(X\geqslant z, Y_i\geqslant X\ \mbox{for all}\ i)=\int_z^{+\infty}\mathrm P(Y_i\geqslant x\ \mbox{for all}\ i)\,f(x)\,\mathrm dx. $$

The simplest proof might be to introduce $Y=\inf_iY_i$. Then $Y$ is a random variable because the family $(Y_i)_i$ is countable. Call $\mu$ the distribution of $Y$. Since $X$ and $Y$ are independent, one has $$ \mathrm P(X\geqslant z, Y_i\geqslant X\ \mbox{for all}\ i)=\mathrm P(Y\geqslant X\geqslant z)=\iint[y\geqslant x\geqslant z]\,f(x)\,\mathrm dx\,\mu(\mathrm dy). $$ Furthermore, for every fixed $x$, $$ \int[y\geqslant x]\,\mu(\mathrm dy)=\mu([x,+\infty))=\mathrm P(Y\geqslant x)=\mathrm P(Y_i\geqslant x\ \mbox{for all}\ i), $$ and the assertion follows.

Application: Choose $X=T_k$, $f(x)=q_k\mathrm e^{-q_kx}$ on $x\geqslant0$, and $(Y_i)_i=(T_j)_{j\ne k}$.

share|cite|improve this answer
In order to obtain the double integral in your answer I proved that, by independence, the distribution of $(X,Y)$ is the product measure of the two distribution. Is this the only/better way obtain the equality? – user7669 Feb 27 '12 at 13:51

I didn't notice the "countable family" part, sorry. Maybe this answer is still of some value.

I will assume $k=1$ here for convenience. The joint distribution function $F(t_2, \dotsc, t_n)$ for $T_2, \dotsc, T_n$ is the product of the individual distribution functions:

$$ F(t_2, \dotsc, t_n) = \prod_{m=2}^nq_m e^{-q_mt_m}. $$

The joint distribution for all variables is therefore

$$ q_1e^{-q_1t_1}F(t_2, \dotsc, t_n). $$

The requested probability is then expressed by the integral

$$ \int_{\Omega}q_1e^{-q_1t_1}F(t_2, \dotsc, t_n) dt_1 \cdots dt_n $$

where $\Omega$ is the set that matches the conditions for this probability:

$$ \Omega = \{ (t_1, \dotsc, t_n) \mid t_1 \geq t \textrm{ and } t_m \geq t_1 \textrm{ for } m = 2, \dotsc, n \} $$

Write the integral more explicitly in terms of the coordinates to get

$$ \int_t^{\infty} \int_{t_1}^{\infty} \cdots \int_{t_1}^{\infty} q_1e^{-q_1t_1}F(t_2, \dotsc, t_n) dt_n \cdots dt_2 dt_1 = \\ \int_t^{\infty} q_1e^{-q_1t_1} \int_{t_1}^{\infty} \cdots \int_{t_1}^{\infty} F(t_2, \dotsc, t_n) dt_n \cdots dt_2 dt_1 = \\ \int_t^{\infty} q_1e^{-q_1t_1} \mathbb{P}(T_m \geq t_1 \textrm{ for } m=2, \dotsc, n) dt_1 $$

share|cite|improve this answer
Note that the OP is interested in a countable collection. – cardinal Feb 24 '12 at 16:18
@cardinal. Oh right, didn't realize that. – WimC Feb 24 '12 at 16:34

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.