A random experiment has exactly three possible outcomes, referred to as outcomes $1, 2,$ and $3,$ with probabilities $p_1 > 0, p_2 > 0,$ and $p_3 > 0,$ where $p_1 +p_2 +p_3 = 1.$ We consider a sequence of independent trials, at each of which the specified random experiment is performed.) For $i = 1, 2,$ let $N_i$ be the number of trials needed for outcome $i$ to occur, and put $N := N_1 \wedge N_2.$

(a) Show that $N$ is independent of $\{N_1 < N_2\}.$

(b) Evaluate $E[N_1 \mid N_1 < N_2].$

(c) Roll a pair of dice until a total of $6$ or $7$ appears. Given that $6$ appears before $7,$ what is the (conditional) expected number of rolls?

The answer to $b)$ is $1/(p_1+p_2)$ and $c)$ is $3.272727$ but I'm unsure of even where to start for $a)$ or the steps involved in arriving at the answers for $b)$ and $c)$

  • $\begingroup$ should I somehow use the fact that 𝐸[1𝐸]=𝑃(𝐸) in the proof? $\endgroup$
    – Hannnnn
    Oct 29, 2019 at 13:43

2 Answers 2


Here is a brute force way to prove (a). I'm sure there is a simpler way to do this, and hopefully someone will post it. We have the joint pdf \begin{align} \mathbb{P}(N_1=k, N_2 = j) &= 1_{k<j} \;p_3^{k-1}p_1(p_1+p_3)^{j-k-1}p_2 + 1_{k>j}\; p_3^{k-1}p_2(p_2+p_3)^{j-k-1}p_1\\ &= p_1p_2p_3^{k-1}\left[ 1_{k<j} \;(p_1+p_3)^{j-k-1} + 1_{k>j}\; (p_2+p_3)^{j-k-1} \right] \end{align} Using the above, one can show that $$ \mathbb{P}(N=n, N_1<N_2 ) = p_3^{n-1}p_1$$ $$ \mathbb{P}(N=n) = p_3^{n-1}(p_1+p_2) $$ $$\mathbb{P}(N_1<N_2 ) = \frac{p_1}{p_1+p_2}$$ And therefore, $$\mathbb{P}(N=n, N_1<N_2 ) = \mathbb{P}(N=n)\cdot \mathbb{P}(N_1<N_2 )$$ proving independence of the events.



$N_i$ is the trial at which $i$ first appears. So $N = \min(N_1, N_2)$ is the trial at which $1$ or $2$ first appears. In other words, before the $N$th trial, only $3$ has been appearing. In other words, $N>n$ is the event that the first $n$ trials are all $3$s.

(a) I would directly calculate $P(N>n \mid N_1 < N_2)$ and $P(N > n \mid N_2 < N_1)$ and show that they are equal. Do you see why this implies the required independence?

(b) Consider $P(N_1 > n \mid N_1 < N_2)$. Can you see why this is exponential?

(c) is just an application of (b)


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .