Given that $R_i$ denotes a random amount (£1 or -£1), depending on the outcome of the coin $i$-th toss:

1) Why $E[R_i R_j]=0$. I can understand this conceptually, I think, or at least I thought I did before I have written that... Does that say expected value of $R_i$ multiplied by$R_j$? And how to show mathematically that it is zero? Would it be: $$\Sigma_{r_i,r_j:p(r_i,r_j)>0} \ r_i r_j p(r_i r_j)$$ No, that can't be true, because then we would have a non-zero probability for at least one combination, and since tosses are independent it would not make sense. Oh.. does it just mean that the tosses are independent?

2) Now if $S_i$ is the total amount of money accumulated up to and including the $i$-th toss, then why is $E[S_i^2]=E[R_1^2+2R_1R_2+...]=i$ ?

3) And a final question. Why is $E[S_6|R_1,R_2,...,R_5]=S_5$

And of course when I say why, I would ideally like to see some extended mathematical explanation.


We have the definition of $E(X)$: $$ E(X) = \sum_{x\in X} xP(x) $$ That is $E(X) = \text{The expected (mean) value when looking at the variable X}$.
And you defined a variable $R_i$ to have the sample space $\{1, -1\}$ (pound). The probability of each sample is the same, because we are flipping coins. So you have $P(1)=0.5,\, P(-1)=0.5$. You also defined the variable $S_i$ to be the accumulated value of $R_j$ from $j = 1$ to $j = i$.

Answer to your first question: $$ E(R_iR_j) = P(r_i=1,r_j=1)\cdot 1\cdot 1 +\\ P(r_i=1,r_j=-1)\cdot 1\cdot (-1) +\\ P(r_i=-1,r_j=1)\cdot (-1)\cdot 1 +\\ P(r_i=-1,r_j=-1)\cdot (-1)\cdot (-1) =\\ \frac{1}{4}\cdot 1\,+\,\frac{1}{4}\cdot (-1)\,+\,\frac{1}{4}\cdot (-1)\,+\,\frac{1}{4}\cdot 1 = 0 $$

Answer to your thrid question: $$ E(S_6)=E(S_5+R_6)=E(S_5)+E(R_6)=S_5+E(R_6)\\\text{where }E(R_6)=1\cdot P(1)+(-1)\cdot P(-1) = 1\cdot 0.5+(-1)\cdot 0.5=0\\\text{So we get } E(S_6)=S_5 $$

The $E(...|...)$ just specifies the variables $R_1$ to $R_5$ is already chosen.

Finally, the hard one can be solved by using a rule for two independent variables: If $X$ is independent on $Y$ then $E(X\cdot Y)=E(X)\cdot E(Y)$. $$ E(S_i^2)=E\left({\left(\sum_{j=1}^i{R_j}\right)}^2\right)= E\left(\sum_{j=1}^i\left(R_j^2+2\sum_{j<k\le i}{R_j\cdot R_k}\right)\right) =\\ \sum_{j=1}^i\left(E(R_j^2)+2\sum_{j<k\le i}{E(R_j\cdot R_k)}\right) =\\ \sum_{j=1}^i\left(E(R_j^2)+2\sum_{j<k\le i}{E(R_j)\cdot E(R_k)}\right) =\\ \sum_{j=1}^i\left(E(R_j^2)+2\sum_{j<k\le i}{0\cdot 0}\right) = \sum_{j=1}^iE(R_j^2) = i\cdot E(R_1^2) = i\cdot\left(0.5\cdot 1^2\, +\, 0.5\cdot (-1)^2\right) = i $$ For those who are interested, this holds too: $$ E(S_i^n)= \begin{cases} 0\quad\;\ n\text{ is odd}\\ i^{\frac n2}\quad n\text{ is even} \end{cases} $$

  • $\begingroup$ This is great! Thank you very much. For the very last line, I do not see how we got $i$ in front of the expected expression, and why did $E[R_j^2]$ change to $E[R_1^2]$? $\endgroup$ – i squared - Keep it Real Nov 4 '15 at 9:14
  • $\begingroup$ We know that $R_j$ does not depend on $j$ (in terms of probability) so we can choose $j$ to be what we want (in this case $1$). $\endgroup$ – Niklas Bäckström Nov 4 '15 at 15:31
  • $\begingroup$ This can be done because the expression inside $E$ does not already contain the variable $R_1$. So an example when this cannot be done would be: $\sum_{j=1}^i{E(R_j\cdot R_1)}$. $\endgroup$ – Niklas Bäckström Nov 4 '15 at 15:41

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.