Throw three dice such that at least two $3$'s shows up with $0.25$ probability

40 Views Asked by At

We consider three fair dice which are thrown simultanously. In other words, one repetition of the experiment means that we have thrown three dice.

Use Chebyshev inequality $P(|X-\mathbb{E}(X)|\geq \epsilon)\leq \frac{\mathbb{V}(X)}{\epsilon^2}$, where $\epsilon>0$, to find the minimum number of repetitions of the experiment such that the probability of getting at least two $3$'s is $0.25$ or more. We can assume that each repitition is independent from others.


I know that this question seems odd because it is easy to do it without Chebyshev inequality. However, our professor wants us to apply Chebyshev inequality.


My approach:

Let be $X_i$ a random variable with $1\leq i\leq n$, such that $X_i=1$, if at least two $3$'s show up and otherwise $X_i=0$. It is easy to see that $P(X_i=1)=\frac{2}{27}$ and $P(X_i=0)=\frac{25}{27}$. Then, we know $Y:=\sum\limits_{i=1}^{n}X_i$ obeys a Binomial distribution and $\mathbb{E}(Y)=\frac{2n}{27}$ and $\mathbb{V}(Y)=\frac{50n}{27^2}$. To answer the question we must find a $n$ such that $P(Y\geq 1)$.

In order to apply Chebyshev inequality we have to do some tricky manipulations and assume that $n$ is large enough such that $\mathbb{E}(Y)-1=\frac{2n}{27}-1>0$.

\begin{align*} &1.) ~P(Y\geq 1)\geq 0.25\iff P(Y<1)\leq 0.75\\ &~\\ &2.)~P(Y<1)=P(Y-\mathbb{E}(Y)\leq 1-\mathbb{E}(Y))=P(\mathbb{E}(Y)-Y\geq \mathbb{E}(Y)-1)\\ &\leq P(|\mathbb{E}(Y)-Y|\geq \mathbb{E}(Y)-1)\leq \frac{\mathbb{V}(Y)}{(\mathbb{E}(Y)-1)^2}\\ &\implies P(Y<1)\leq \frac{\mathbb{V}(Y)}{(\mathbb{E}(Y)-1)^2}=\frac{\frac{50n}{27^2}}{\left(\frac{2n}{27}-1\right)^2}. \end{align*}

In equation 2.) I have used the subset property and applied Chebyshev inequality. Long story short, we are looking for a $n\in\mathbb{N}$ such that $\frac{\frac{50n}{27^2}}{\left(\frac{2n}{27}-1\right)^2}\leq 0.75$. After some manipulations of the inequality we see that $n\geq 39$.


My tutor deducted almost $50$% of the points just mentioning that I can't do it this way and it is not clear if the $n$ I found is truely the smallest number of repitition if we use Chebyshev inequality. To be honest I don't know what she meant and I don't see any mistakes. Maybe someone else can explain to me what is wrong with my solution.