0

I'm trying to approach this question using generating functions.

I set the problem up similar to a "toss $17$ balls into $9$ bins, what's the probability that no bin gets $6$ or balls in it." as the $9$ bins correlate to the $9$ possible places between tails that heads can occur in.

My generating function ended up like this:

$$(1-x^6)^9 \times (1/(1-x))^9 = 17$$

and this is where I sort of hit a dead end in my work. I assume we should solve for coefficients of $f(x)g(x)$, making sure they add up to $17$.

Starting this, I found that the only coefficients that were nonzero and under $17$ for $(1-x^6)^9$ would be $0, 6$, and $12$. But these are giving me some ugly numbers and I don't really know where to go from here, or how to calculate the probability at the end.

Could anybody give me some pointers on where to go from here?

N. F. Taussig
  • 76,571
  • 1
    There is a great answer given by gAr over here. http://math.stackexchange.com/questions/846442/generating-function-probability-regarding-coin-toss/846666#846666 – bobbym May 01 '15 at 04:19

2 Answers2

1

Your approach using generating functions is correct. I will work the whole solution so others still trying to understand generating functions can follow.

First, we know the probability will be the following occurrences,

( 8 tails and 17 heads with runs < 6)/( 8 tails and 17 heads)

To calculate the function occurrences without runs visualize it as a boxes and balls problem. Separate the boxes with the 8 Tails

_T_T_T_T_T_T_T_T_

This leads to 8 boxes and 17 balls to place. (25 - 8 = 17)

Since our boxes can contain 0-5 balls the function takes the form for a single box

$x^0+x^1+x^2+x^3+x^4+x^5=\mathbf{1+x^1+x^2+x^3+x^4+x^5}$

Since we have 9 boxes, we take this result to the 9th power $\mathbf{(1+x^1+x^2+x^3+x^4+x^5)^9}$

using $\frac{1-x^{m+1}}{1-x} = 1+x+x^2+...+x^m $ $$\mathbf{1+x^1+x^2+x^3+x^4+x^5 =\frac{(1-x^6)^9}{(1-x)^9}}$$

Break this into two functions f(x) and g(x)

$$f(x)= (1-x^6)^9$$ $$f(x)= \frac{1}{(1-x)^9}$$

Their corresponding expansions are,

$(1-x^m)^n=1-\binom{n}{1}x^{1*m}+\binom{n}{2}x^{2*m}....+(-1)^n\binom{n}{n}x^{nm}$

$\frac{1}{(1-x)^n}=1+(\frac{1+n-1}{1})x+(\frac{2+n-1}{2})x^2+......+(\frac{r+n-1}{r})x^r$

We need to find the coefficient for $x^{17}$ which is equal to $$\Sigma f(x)g(x)$$ Where the resulting power of x is 17

The only values of $\mathbf{f(x)}$ that can meet this constraint are $\mathbf{x^0,x^6,x^{12}}$

Lets represent the power of x in f(x) with $a^k$ and g(x) with $b^k$ This leads to the only valid combinations being $$a^0b^{17}+a^6b^{11}+a^{12}b^5$$

Therefore, the number of occurrences of 8 tails and 17 heads with runs < 6 is $$1*\binom{17+9-1}{17}-\binom{9}{1}\binom{11+9-1}{11}+\binom{9}{2}\binom{5+9-1}{5}$$

This evaluates to 447669 occurrences.

Next we find all possible occurrences with no constraints. Using the formula for placement of n identical objects in k boxes $\binom{n+k-1}{n}$ Which gives us $$\binom{17+9-1}{17}=1081575$$

$$100*\frac{447669}{1081575}\approx41.39\%$$

CoolCat
  • 11
  • 1
0

I just ran a simple computer simulation and got $447,669$ "winners" out of $1,081,575$ cases which have exactly $8$ tails occurring (out of $25$ fair coin flips), so the probability is $41.39$%. I looped thru all $2^{25}$ = $33,554,432$ possible states of $25$ coin flips.

David
  • 1,702