We are tossing 2 fair coins once. Then for the next toss we are only tossing the coin(s) that came up heads before.

Let $X$ be the total number of heads.

The question is $EX$ and the distribution of $X$

I tried to calculate the expected values for small $X=x$-es but it gets really complicated soon.

  • 1
    $\begingroup$ Perhaps you should include what you have calculated thus far so that people will know exactly where you are stuck or made a wrong turn. Also, you might want to start with something smaller. If $Y$ is the number of Heads on the first toss of 2 fair coins, what is the distribution of $Y$ and what is $E[Y]$? $\endgroup$ – Dilip Sarwate Oct 12 '14 at 19:25
  • $\begingroup$ $TT, THT, THH, HTT, HTH, HHTT, HHHT, HHTH, HHHH$ $\endgroup$ – Dilip Sarwate Oct 12 '14 at 21:11
  • 1
    $\begingroup$ Can you show some of your work? It might help to guide you. $\endgroup$ – Glen_b Oct 12 '14 at 23:35
  • $\begingroup$ What exactly do you mean by "small $X$"? It sounds like you will make at most two tosses and can observe at most four total heads; there do not seem to be any complications insofar as that goes. Thus the reference to "small $X$" suggests that maybe you are contemplating a different interpretation of the problem which somehow did not get expressed explicitly. $\endgroup$ – whuber Oct 13 '14 at 4:44
  • $\begingroup$ @whuber maybe my interpretation of the problem is indeed incorrect. I thought we toss a coin that came up heads until we got a tails, so it is not just a 2 round game. $\endgroup$ – kanbhold Oct 13 '14 at 5:38

Start by considering each coin in isolation, and the question becomes easier. Let $Y$ denote the number of heads for the first coin, and $Z$ denote the number of heads for the second coin, so $X=Y+Z$. $Y$ and $Z$ are identically distributed so let's just consider $Y$.

First, we know that if $Y=y$, then the first coin must have come up heads exactly $y$ times, and then tails once. The probability of $y$ heads in a row is $\left ( \frac{1}{2} \right )^{y}$, and the probability of getting tails after that is $\frac{1}{2} $. Thus:

$P(Y=y) = \left ( \frac{1}{2} \right )^{(y+1)}$

To calculate the expected value of $Y$, we sum $y \cdot P(Y=y)$ over all values of $Y$, from zero to infinity:

\begin{align} E[Y] &= \sum_{y=0}^{\infty} y \cdot P(Y = y)\\ &= \sum_{y=0}^{\infty} y \cdot \left ( \frac{1}{2} \right )^{(y+1)} \\ &= 1 \end{align}

The expectation of the sum of two random variables is the sum of their expectations, so $E[X] = E[Y+Z] = E[Y] + E[Z] = 2$.

How do we use $P(Y)$ to get $P(X)$? Here's an example: Say $X=2$. Then we know there are three possibilities: (1) $Y=2$ and $Z=0$, (2) $Y=1$ and $Z=1$, or (3) $Y=0$ and $Z=1$. Since $Y$ and $Z$ are independent, we have: \begin{align} P(X=2) &= \left( \left( \frac{1}{2} \right)^{3}\cdot \frac{1}{2} \right ) + \left( \left( \frac{1}{2} \right)^{2}\cdot \left(\frac{1}{2} \right)^2 \right) + \left( \frac{1}{2} \cdot \left( \frac{1}{2} \right)^{3} \right )\\ &= 3 \cdot \left( \frac{1}{2} \right)^{4} \end{align}

This example gives the intuition that maybe $P(X=x) = (x+1) \cdot \left( \frac{1}{2} \right)^{(x+2)}$. It is true for $X=0$: both heads have to come up tails on the first flip, and the probability of that occurring is $\frac{1}{4} = (0+1) \cdot \left( \frac{1}{2} \right)^{(0+2)}$.

It should be simple to show by induction that this is true for all values of $X$. Here is a sketch. First note that if $X=x$ there are $x+1$ possible combinations of $Y$ and $Z$ values that can produce $y + z = x$. Each value of $Y$ corresponds to a unique series of heads and a tail (and likewise for $Z$). If we iterate, and ask what values of $Y$ and $Z$ could give $y' + z' = x+1$, we can start with our original set of possible combinations of $Y$ and $Z$ values, and just add an extra head to the start of each run for the first coin, which would multiply the probability of each combination by $\frac{1}{2}$. That is, we set $y'= y+1$ and $z'=z$. Then we need to add one new term to the sum, to account for the case where $y=0$, and $z'=x+1$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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