I have three independent random variables, $X_1$, $X_2$, and $X_3$, and each random variable has different mean. (And I assume they have the same variance). I would like to know how I can get the probability that $X_1$ is the max of all three. And help would be appreciated.

  • $\begingroup$ Do you know the distributions? $\endgroup$ Commented May 16, 2016 at 18:23
  • $\begingroup$ I assume they all have the normal distributions with same variance but different means. $\endgroup$
    – Jae Chung
    Commented May 16, 2016 at 18:29
  • $\begingroup$ Do values of X1 correspond to values of X2 & X3? $\endgroup$ Commented May 16, 2016 at 18:30
  • $\begingroup$ gung, I don't understand your question. Could you clarify? $\endgroup$
    – Jae Chung
    Commented May 16, 2016 at 18:39
  • $\begingroup$ Closely related: stats.stackexchange.com/questions/87241 (generic formulation), stats.stackexchange.com/questions/173064 (two independent normals), stats.stackexchange.com/questions/139072 (two correlated normals), etc. Note that $X_2-X_1$ and $X_3-X_1$ are correlated Normals and $X_1$ is the max of all $X_i$ if and only if the larger of these two differences is non-positive. Thus, you may directly apply the answer in the last link. $\endgroup$
    – whuber
    Commented May 16, 2016 at 20:04

1 Answer 1


Assume their means are $\mu_1,\mu_2,\mu_3$, with the same standard deviation, $\sigma$. Then the probability that the first random variable is the max is:

$$P(X_1>X_2,X_1>X_3)=\int_{-\infty}^\infty P(x>X_2|X_1=x)P(x>X_3|X_1=x)f(x)dx$$

where $f(x)$ is the density of $X_1$. It's doubtful whether you can simply this further.

  • $\begingroup$ Since the variables are independent this simplifies to $\int_{-\infty}^{\infty} {P(X_2 < x) P(X_3 < x) f(x) dx}$ $\endgroup$
    – jodag
    Commented Apr 24, 2023 at 1:43

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.