Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

I'm doing a branching process problem and am not sure I did it correctly. Suppose $X_0 = 1$ and $p_0 = .5, p_1 = .1,$ and $p_3 = .4$ represent the probabilities that zero, one, and three individuals will be produced, respectively. There are two parts to the question:

a) Find $P(X_2 = 0 \mid X_1 > 0)$

b) Find $P(X_3 = 0 \mid X_2 > 0)$.

I imagine the solutions to these problems are very similar.

To solve part a), I first found the expected number of individuals produced by the zeroth generation to be $1.3$. Next, I found the extinction probability $a$ by solving the equation

$$a = .5 + .1a + .4a^3,$$

where $a$ came out to be roughly $0.725$. All I could think of doing was to solve for $$P(X_2 = 0 \mid X_1 = E[X_1]),$$ where $$E[X_1] = 1.3.$$ To solve this, I know that you can simply calculate $a^{1.3} = .658$. However, this doesn't feel like it's the right approach since it doesn't cover all possible values of $X_1$; it only takes into account the expected value at time $t = 1$. I figure one could correct this error by solving

$$P(X_2 = 0 \mid X_1 > 0) = \frac{P(X_2 = 0 \cap X_1 > 0)}{P(X_1 > 0)} $$

by Bayes' Theorem, but unfortunately I don't know how one would go about computing that. Is anything I have here at least remotely correct, or am I just throwing out jibber-jabber?

Side note: For part b), I raised the extinction probability to the expected number of individuals are time $t = 2$ $(E[X_2] = 2.6)$ and got $a^{2.6} = .433$. Again, I'm not so sure this is correct, but it was my best attempt.


share|cite|improve this question
up vote 2 down vote accepted

Given that $X_1>0$ there can be $1$ or $3$ children on generation $1$, so you could use conditioning on the number of children on generation $1$ to get $a)$. This would lead to $$ \begin{align} P(X_2=0)&=P(X_2=0|X_1=0)P(X_1=0)+P(X_2=0|X_1=1)P(X_1=1)+P(X_2=0|X_1=3)P(X_1=3)\\[5pt] &= \frac{1}{2}+\frac{1}{2}\cdot\frac{1}{10}+\frac{1}{2^3}\cdot\frac{4}{10}\\[5pt] &=\frac{3}{5}. \end{align} $$

Now you can rewrite this as $$ \begin{align*} P(X_2=0)&=P(X_2=0\,|\,X_1=0)P(X_1=0)+P(X_2=0\,|\,X_1>0)P(X_1>0)\\[5pt] \frac{3}{5}&=\frac{1}{2}+P(X_2=0\,|\,X_1>0)\frac{1}{2} \end{align*} $$ which you can isolate to find $P(X_2=0\,|\,X_1>0)=\frac{1}{5}$.

You can do the same for $b)$, it only takes a few more cases.

share|cite|improve this answer
You've calculated $P(X_2=0, X_1>0)$ not $P(X_2=0\,|\, X_1>0)$. – Byron Schmuland Sep 25 '12 at 13:44
You were right thanks! Fixed it. – Jean-Sébastien Sep 25 '12 at 15:32
This is great, thanks! – jrad Oct 10 '12 at 6:03

The function $\phi(t)={1\over 2}+{t\over 10}+{2 t^3\over 5}$ will prove very useful, since $\mathbb{P}(X_n=0)$ is equal to $\phi(\phi(\cdots(\phi(0))\cdots))$ where there are $n$ applications of the function $\phi$. Then you can easily calculate quantities like $$\mathbb{P}(X_2=0, X_1>0)=\mathbb{P}(X_2=0)-\mathbb{P}(X_1=0).$$ Can you take it from here?

share|cite|improve this answer
This all makes sense up until you write up your equation. I can see how to apply it, but I'm not seeing why $$P(X_2 = 0, X_1 > 0) = P(X_2 = 0) - P(X_1 = 0).$$ Could you explain that? – jrad Oct 10 '12 at 6:01
Scratch that--A little consideration of my own helped me understand it. Thanks! – jrad Oct 10 '12 at 6:02

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.