0

I want to do coupon collector's for a dice roll, or the expected number of rolls to get all 6 numbers from definition of expected value as opposed to using linearity of expectation.

$\tau = min\{ t \vert X_t = i, \forall k \neq i,\ \exists j<t: X_j = k ,\ \forall l< t, \ X_l \neq i \}$.

$\mathbb{E}[\tau] = \sum_{j = 1}^{\infty} j\mathbb{P}(\tau = j)$ is the definition.

I try to compute $\mathbb{P}(\tau = j)$, and for j rolls, you must have 5 numbers in the first $j-1$ rolls, and a 6th number in the $j$th spot, and the 6th number can't be in any of the first $j-1$ spots. So

$\mathbb{P}(\tau = j) = 6[\frac{1}{6^6}5! \binom{j-1}{5} (\frac{5}{6})^{j-6}$].

$\frac{1}{6^6}$ is for six numbers being in those spots (say, 1-5 being in the first j-1 spots, 6 being in the last), $5!$ is the ways of arranging the 5 numbers, $\binom{j-1}{5}$ is to choose the ways of putting $1,2,3,4,5$ in $j-1$ spots, and $(\frac{5}{6})^{j-6}$ is so the $j-1$ spots which you didn't place 1 to 5, don't contain a 6. then multiply by 6 since 1 through 5 can also be at the jth position.

This doesn't work out to be the right sum. Where is this wrong?

Wanshan
  • 1,343
  • 10
  • 19
blanchey
  • 801
  • 2
    Computing $P(\tau = j)$ is not as easy as you make out. Having singled out the $6^{th}$ face that will appear on trial $j$ you have to guarantee that all $5$ of the other faces appear in the first $j-1$ trials. That, of course, is more or less the problem you set out to solve (only with $5$ faces instead of $6$). – lulu May 24 '18 at 16:02
  • The order of your quantifiers is reversed; you would like for every $k\neq i$ that there exists a $j$ with $X_j=k$. The way it's written now implies that the same $j$ should work for each $k$. – rikhavshah May 24 '18 at 16:05
  • 1
    First try computing $\Pr[\tau > j]$, i.e. the probability that in $j$ rolls one of the six faces is missed. Then $\Pr[\tau=j] = \Pr[\tau > j-1]-\Pr[\tau > j]$ – rikhavshah May 24 '18 at 16:08

0 Answers0