3
$\begingroup$

Stuck with this challenge my professor gives us. Does anyone have a good example? I've seen examples for two variables, not for three though. Here is the full problem:

Give example of Bernoulli random variables $X,Y,Z$ which are not independent, $E[X]=E[Y]=E[Z]=1/2$, and $E[XYZ]=E[X]·E[Y]·E[Z]$.

$\endgroup$
2
  • $\begingroup$ Could you illustrate the 2 variable problem? It could provide some useful insights for the 3 variable problem. Also, since this is homework, please tag self-study $\endgroup$
    – jcken
    Mar 9, 2021 at 8:00
  • $\begingroup$ It is not the case for only two Bernouilli random variables. See physicsforums.com/threads/… or see that if following the method in my answer that there would be a consistent linear system of four equations and only four free variables. $\endgroup$
    – svendvn
    Mar 9, 2021 at 10:09

1 Answer 1

4
$\begingroup$

The requirement $E[XYZ]=E[X]E[Y]E[Z]$ is the same as $P(X=1, Y=1, Z=1)=0.125$. Given the other requirements, $P(X=1)=0.5$, $P(Y=1)=0.5$ and $P(Z=1)=0.5$ (and the fact that the total probability should add up to 1) we can establish a set of equations. Let $P(X=x,Y=y,Z=z)=p_{x,y,z}$ and we can write

$$ p_{1,1,1}=0.125\\ p_{1,0,0}+p_{1,0,1}+p_{1,1,0}+p_{1,1,1}=0.5\\ p_{0,1,0}+p_{0,1,1}+p_{1,1,0}+p_{1,1,1}=0.5\\ p_{0,0,1}+p_{0,1,1}+p_{1,0,1}+p_{1,1,1}=0.5\\ p_{0,0,0}+p_{0,0,1}+p_{0,1,0}+p_{1,0,0}+p_{0,1,1}+p_{1,0,1}+p_{0,1,1}+p_{1,1,1}=1 $$ To find an example, we have to find a solution to the above set of equations,that satisfy that all $p$'s are between 0 and 1, but for which at least one of the $p_{x,y,z}\neq 0.125$. And after looking a little I found the solution:

$$ p_{1,1,1}=1/8\\ p_{1,0,0}=p_{0,1,1}=1/4,\\ p_{0,1,0}=p_{0,0,1}=p_{1,1,0}=p_{1,0,1}=1/16\\ p_{0,0,0}=1/8. $$

$\endgroup$
5
  • $\begingroup$ The solution is simpler than you suggest, because the required condition only is that $\Pr(X=1,Y=1,Z=1)=E[XYZ]=E[X]E[Y]E[Z]=(1/2)^3.$ Thus, all you have to do is distribute the remaining probability of $1-1/2^3 = 7/8$ among the other seven possible outcomes in an uneven way. Indeed, if you did this randomly there is a 100% chance that $(X,Y,Z)$ would not be independent! $\endgroup$
    – whuber
    Mar 9, 2021 at 16:00
  • 2
    $\begingroup$ @whuber But what you suggest would not satisfy the requirement that $EX=EY=EX=1/2$ . $\endgroup$ Mar 9, 2021 at 16:07
  • $\begingroup$ While I can see this is clearly a solid answer, my professor was more looking for a specific/real world example. I am still struggling to come up with something that might model this behavior. Does anyone have any suggestions? $\endgroup$
    – Ralph
    Mar 9, 2021 at 21:58
  • 1
    $\begingroup$ Since all probabilities in the example are of the form 1/2^n, n<5 it is possible to describe it with four coin tosses, but I am not sure how natural it will seem; for example let X Y and Z be coin tosses. If all three are heads or all three are tails, do nothing. If they are different, throw another coin. If the other coin says head turn Y and Z to the opposite of X. This should produce the probabilities above $\endgroup$
    – svendvn
    Mar 9, 2021 at 23:38
  • 1
    $\begingroup$ @Jarle Thank you, I overlooked that expectation requirement, so that rules out the random solution. $\endgroup$
    – whuber
    Mar 10, 2021 at 15:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.