2
$\begingroup$

Assuming that I roll 10 dice, and am going to call a 5 or 6 a 'success' on any single die, how can I determine the average number of successes, the variance, and the standard deviation for the associated bell curve?

If I use $N$ as the number of dice used, and $P$ as the probability of getting a success on a single die roll, I believe the average is simply $NP$, or in this case, $10×\frac26=3.33$ successes on average.

What I'd like to know now, is what is my range of possible successes at 1 and 2 standard deviations. Beyond that, I'd like to understand how to perform this calculation assuming I change either $N$ or $P$.

Thanks in advance!

$\endgroup$
  • $\begingroup$ In my opintion 10 dice (random variables) are not enough to approximate the distribution by the normal distribution. The thumb rule is $n\cdot p\cdot q > 9$. Think about that. $\endgroup$ – callculus Dec 6 '17 at 1:32
0
$\begingroup$

The distribution is binomial with p=1/3. Use the standard form to get the first and second moment and therefore the variance and standard deviation.

$\endgroup$
0
$\begingroup$

As mentioned by callculus, normal approximation would not be a good method in this case. Instead I will use the binomial distribution.

The expected value can be obtained in two ways:

$$\begin{align*} E(X) &=\sum_{k=0}^ n {n\choose{k}} k\cdot p^k\left(1-p\right)^{n-k}\\\\ &=\sum_{k=0}^{10} {10\choose{k}} k\cdot \left(\frac{1}{3}\right)^k\left(1-\left(\frac{1}{3}\right)\right)^{10-k} \\\\ &= \frac{10}{3}\\\\ &\approx 3.333 \end{align*}$$

You could also just do $np$ to get $$E(X)=10\cdot\frac{1}{3}=\frac{10}{3}\approx 3.333$$

The variance can be obtained in two ways:

$$\begin{align*} Var(X) &=E(X^2)-E(X)^2 \\\\ &=\sum_{k=0}^{10} {n\choose{k}} \cdot k^2 \cdot p^k\left(1-p\right)^{n-k} - \left(\sum_{n=0}^{10}{n\choose{k}} \cdot k \cdot p^k\left(1-p\right)^{n-k}\right)^2 \\\\ &=\sum_{k=0}^{10} {10\choose{k}}\cdot k^2 \cdot \left(\frac{1}{3}\right)^k\left(1-\left(\frac{1}{3}\right)\right)^{10-k} - \left(\sum_{n=0}^{10} {10\choose{k}}\cdot k \cdot \left(\frac{1}{3}\right)^k\left(1-\left(\frac{1}{3}\right)\right)^{10-k}\right)^2 \\\\ &= \frac{40}{3} - \left({\frac{10}{3}}\right)^2\\\\ &= \frac{20}{9}\\\\ &\approx 2.222 \end{align*}$$

You could also just do $npq$ to get

$$Var(X)=10\cdot\frac{1}{3}\cdot\frac{2}{3}=\frac{20}{9}\approx 2.222$$

The standard deviation is just

$$S.D.=\sqrt{Var(X)}=\sqrt{\frac{20}{9}}\approx 1.490711985$$

$\endgroup$
  • 1
    $\begingroup$ @Calculus and Remy - Thank you both very much. I had come up with something similar to what Remy posted, and it seemed to be working fine, but when I tried the equation with 100 dice, it fell apart. I wasn't considering the NPQ>9 thumb rule. Thanks again! $\endgroup$ – JJ Bone Dec 6 '17 at 12:39

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.