Exercise 2.17 - Expected value of the minimum

Answers

Let m denote the location of the left most point, we have:

p ( m > t ) = p ( [ X > t ]  and  [ Y > t ] ) = p ( X > t ) p ( Y > t ) = ( 1 t ) 2 .

Therefore:

𝔼 [ m ] = 0 1 t p ( m = t ) d t = 0 1 p ( m > t ) d t = 0 1 ( 1 t ) 2 d t = 1 3 .

For the second equation, note that:

p ( m t ) = t 1 t p ( m = t ) d t ,

therefore

0 1 t 1 t p ( m = t ) d t d t = 0 1 0 t p ( m = t ) d t d t = 0 1 t p ( m = t ) d t .

There are perhaps more intuitive solutions to this problem, for example, plotting the value of X , Y and min ( X , Y ) into one graph:

Figure 1: Exercise 2.17, P1.

The height of the cyan pyramid at ( x , y ) marks the value of min ( x , y ) , so the expectation of the statistics equals the average height, in this case also the volume of the pyramid, 1 3 . One can also graphically compute the average distance between X and Y from the following plot:

Figure 2: Exercise 2.17, P2

Since the average distance between X and Y is 1 3 , so is that between min ( X , Y ) and max ( X , Y ) . Moreover, we have 𝔼 [ min ( X , Y ) + max ( X , Y ) ] = 𝔼 [ X ] + 𝔼 [ Y ] = 1 , hence min ( X , Y ) = 1 3 .

However, the graphical method, although entertaining and inspiring, should not be considered as a reliable option in proving probability properties. The dependency can complicate the underlying topology (e.g., the X Y plane might have another geometry other than the Euclidean one, if X and Y are not independent), resulting in confusions and fallacies.

For example, if X is subject to a uniform distribution on [ 0 , 1 ] while Y is uniformly distributed in [ max ( 0 , X 0.2 ) , min ( 1 , X + 0.2 ) ] . Then the pyramid in Fig. 1 is left with the region along the diagonal line in the X - Y plane. This generalization is insignificant since it only changes the region where the integral shall be done.

Consider the case where X and Y are independent random variables subject to truncated Gaussian centered at 1 2 on [ 0 , 1 ] . The plot for visualizing min ( X , Y ) is the same as Fig. 1. However, to compute the expectation of min ( X , Y ) , one cannot simply calculate the volume of the pyramid since the geometry of the X - Y plane has changed.

User profile picture
2021-03-24 13:42
Comments