[00:00] Say you sample two random numbers, each one uniform in the range from 0 to 1, Then the result is of course another random number with this bias towards being larger.
[00:14] be to take one of those numbers and compute its square root. When you square a number that's smaller than 1, it becomes smaller, So this is another process that would give you
[00:29] The surprise is that both of these are the same, in the sense that the distribution describing your result is identical for At first it just feels really wrong that computing a maximum and
[00:43] but there's actually a really nice way to visualize why this should be true. Well think about one of these random numbers as existing somewhere on an x-axis between
[00:55] And the other random number is going to exist on a y-axis, again uniform between 0 and 1. when you sample both at random, you're basically sampling a random point inside this
[01:08] So take a moment to think about what it looks like for the maximum of these two values to be a particular number, like 0.7. Well either x1 is equal to that value and x2 is smaller than it,
[01:22] which puts you somewhere on this line, or x2 equals that value and x1 is smaller than that, putting you somewhere on this line. In general, with continuous values, it's not very helpful to ask the probability
[01:35] of equaling a certain number, since the answer tends to be infinitesimal. random value is less than or equal to a certain number. In this case, what it looks like to be less than or equal to 0.7 is that
[01:50] you fall somewhere inside this square here, and so because everything is uniform, the probability of landing in that region is the area of that region. is less than some number r looks like r-squared.
[02:06] cumulative distribution function for the random variable. What is the probability that the square root of one of these values is less than some number r?
[02:21] That's the same thing as asking for the value itself to be less than or equal to r-squared, and since it's all uniform, the answer there is again r-squared. So both of these processes have the same cumulative distribution function,
[02:34] Essentially, identical reasoning will show that if you take the maximum of three such random variables, it has the same effect as taking the cube root of one of them.