Median of Square's Area

Calculus Level 4

The length ( L ) (L) and width ( W ) (W) of a rectangle are random variables, independently and uniformly distributed on the interval [ 9 , 11 ] , [9,11], and the area ( A ) (A) of the rectangle is also a random variable: L , W U ( 9 , 11 ) ; A = L W . L,W \sim U(9,11);\ \ \ \ A = LW. It is not difficult to see that the expected value of the area is E [ A ] = 100. E[A] = 100.

Determine the median of the area.


The answer is 99.6773.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

5 solutions

天则 空想
Jan 4, 2018

Shaded areas represent possible values for (L, W). If you can find a curve LW = A bisects the area of this square, then A is the required median. Calculate the shadow area of the upper half, we come to this formula. The final solution A = 99.6773

This is the best and most elegant answer. (I am biased because I did it the same way!)

Patrick Corn - 3 years, 5 months ago

Same answer!

Kevin Tong - 3 years, 5 months ago
Arjen Vreugdenhil
Dec 12, 2017

We will determine the probability function F A ( a ) = P ( A < a ) F_A(a) = \mathbb P(A < a) . The median is then the value m m for which F A ( m ) = 0.5 F_A(m) = 0.5 .

The distributions for W W and L L have probability function F W , L ( x ) = { 0 x < 9 1 2 ( x 9 ) 9 x 11 1 x > 11 f W , L ( x ) = F W , L ( x ) = { 0 x < 9 1 2 9 x 11 0 x > 11 F_{W,L}(x) = \begin{cases} 0 & x < 9 \\ \tfrac12(x-9) & 9 \leq x \leq 11 \\ 1 & x > 11\end{cases} \ \ \ \ \ f_{W,L}(x) = F'_{W,L}(x) = \begin{cases} 0 & x < 9 \\ \tfrac12 & 9 \leq x \leq 11 \\ 0 & x > 11 \end{cases} In principle, F A ( a ) = P ( A < a ) = w = a f W ( w ) f L ( ) d w d = 9 11 1 2 F L ( a / w ) d w . F_A(a) = \mathbb P(A < a) = \iint_{\ell w = a} f_W(w)f_L(\ell)\:dw\:d\ell = \int_9^{11} \tfrac12 F_L(a/w)\:dw. If a < 81 a < 81 , we have a / w < 9 a/w < 9 on the entire integration interval, so that F A ( a ) = 0 F_A(a) = 0 .

If a > 121 a > 121 , we have a / w > 11 a/w > 11 on the entire integration interval, so that F A ( a ) = 9 11 1 2 1 d w = 1 F_A(a) = \int_9^{11} \tfrac12\cdot 1\:dw = 1 .

If 81 a 99 81 \leq a \leq 99 , then a / w < 11 a/w < 11 on the entire integration interval, but a / w < 9 a/w < 9 for w > a / 9 w > a/9 . Therefore we have F A ( a ) = 9 a / 9 1 4 ( a w 9 ) d w = 1 4 [ a ln w 9 w ] 9 a / 9 = 1 4 ( a ln a 2 a ln 9 + 81 a ) . F_A(a) = \int_9^{a/9} \tfrac14\left(\frac a w - 9\right)\:dw = \tfrac14\left[a\ln w - 9w\right]_9^{a/9} = \tfrac14\left(a\ln a - 2a \ln 9 + 81 - a \right).

If 99 a 121 99 \leq a \leq 121 , then a / w > 9 a/w > 9 on the entire integration interval, but a / w > 11 a/w > 11 for w < a / 11 w < a/11 . Therefore we have F A ( a ) = 1 a / 11 11 1 4 ( 11 a w ) d w = 1 1 4 [ 11 w a ln w ] a / 11 11 = 1 4 ( 2 a ln 11 a ln a + 117 a ) . F_A(a) = 1 - \int_{a/11}^{11} \tfrac14\left(11 - \frac a w\right)\:dw = 1 - \tfrac14\left[11w - a\ln w\right]_{a/11}^{11} = \tfrac14\left(2a\ln 11 - a\ln a + 117 - a \right).

Thus the probability function for the area is F A ( a ) = { 0 a < 81 1 4 ( a ln a 2 a ln 9 + 81 a ) 81 a 99 1 4 ( 2 a ln 11 a ln a + 117 a ) 99 a 121 1 a > 121 F_A(a) = \begin{cases} 0 & a < 81 \\ \tfrac14\left(a\ln a - 2a \ln 9 + 81 - a\right) & 81 \leq a \leq 99 \\ \tfrac14\left(2a\ln 11 - a\ln a + 117 - a\right) & 99 \leq a \leq 121 \\ 1 & a > 121\end{cases} At the transition point F A ( 99 ) 0.466600 F_A(99) \approx 0.466600 ; the median is therefore greater than 99, and we solve numerically 1 4 ( 2 a ln 11 a ln a + 117 a ) = 0.5 2 a ln 11 a ln a + a = 119 m = a 99.6773 . \tfrac14\left(2a\ln 11 - a\ln a + 117 - a \right) = 0.5\ \ \ 2a\ln 11 - a\ln a + a = 119\ \ \ \therefore\ \ \ m = a \approx \boxed{99.6773}.


Graphs of the probability function F A F_A and probability density function f A = F A f_A = F'_A :

How is the median of a random variable defined?

Agnishom Chattopadhyay - 3 years, 5 months ago

Log in to reply

Given a random variable X X , determine x x_- and x + x_+ such that x = sup { P ( X x ) < 1 2 } ; x + = inf { P ( X x ) < 1 2 } . x_- = \text{sup}\ \{ \mathbb P(X \leq x) < \tfrac12 \}; \\ x_+ = \text{inf}\ \{ \mathbb P(X \geq x) < \tfrac12 \}. The median is the mean of these two values, med X = 1 2 ( x + x + ) \text{med}\ X = \tfrac12(x_- + x_+) .

In many cases, it is simply the unique value x x for which P ( X x ) = 1 2 \mathbb P(X \leq x) = \tfrac12 , i.e. F X ( x ) = 1 2 F_X(x) = \tfrac12 . The technical details in the definition are necessary if there is no such unique value.

Arjen Vreugdenhil - 3 years, 5 months ago

I have a question. Why do you divide by 4 in the integral 9 a 9 1 4 ( a w 9 ) d w \int_9^{\frac{a}{9}}\frac{1}{4}\left(\frac{a}{w}-9\right)\ dw when the length of the interval is 11 9 = 2 11-9=2 ?

Vishaal Ram - 3 years, 5 months ago

Log in to reply

The first factor 1 2 \tfrac12 comes from the probability function for the length, F ( ) = 1 2 ( 9 ) F(\ell) = \tfrac12(\ell - 9) . The second factor 1 2 \tfrac12 comes from the probability density for the width, f ( w ) = 1 2 f(w) = \tfrac12 .

Or, the random variable vector ( L , W ) (L,W) is uniformly distributed over [ 9 , 11 ] × [ 9 , 11 ] [9,11]\times[9,11] , a square of area four. In order for the total probability to be one, we need d P ( , w ) = 1 4 d d w dP(\ell,w) = \tfrac14\:d\ell\:dw .

Arjen Vreugdenhil - 3 years, 5 months ago

Log in to reply

I see. Thank you.

Vishaal Ram - 3 years, 5 months ago

Also, I think you made a little mistake while integrating. For example, the first part of the CDF I think should be 1 4 ( a ln ( a ) 2 a ln 9 a + 81 ) \frac{1}{4}\left(a\ln\left(a\right)-2a\ln9-a+81\right)

Vishaal Ram - 3 years, 5 months ago
Vishal Gokhale
Jan 2, 2018

Mean (81,121) = 101 this is wrong answer .. but the app allows it

You should mention this in the "Reports" section, not the solutions section.

Brian Egedy - 3 years, 5 months ago
Jacki Kay
Jan 2, 2018

For each value, we have 3 possible values 9, 10 and 11 As such, the possible outcomes for A are 81 90 99 100 110 121. everything but the squares occur 2 times {90, 99, 110} while the squares occur 1 time {81, 100, 121}.

Therefore the complete set looks like this:

{81, 90, 90, 99, 99, 100, 110, 110, 121} which includes

1 2 3 4 5 6 7 8 9

The result should, therefore, be 99. Even if I made a mistake, the only possible answers could ever be 99 or 99,5 as in (99+100)/2. Whatever was calculated was NOT the mean! It probably was the average, but I'm too lazy to check. Just as I am too lazy to format this here, even if it wouldn't be hard in Latex...

I believe this problem includes all the values between 9 and 11 (eg 9.839385 and 10.40872134) so it isn't as simple as that.

Dylan Wedel - 3 years, 5 months ago

Log in to reply

That is correct. There is no reason to assume that the length or width should be an integer.

You approach is in principle correct. However, there are an infinite number of possible values for the area, which makes taking the median a lot harder. My solution basically boils down to find out "how many times" each of the possible values of the area (anything from 81 to 121) occurs, and picking the middle value of this weighted set of values. But because of the infinite number of possibilities, we need the theory of continuous probability distributions and integral calculus.

An intermediate approach would be as follows: in the interval [ 9 , 11 ] [9,11] , pick (say) 20 equidistance points. Calculate the area for each of these cases. Sort them from low to high and choose the mean of the middle elements. Here is a spreadsheet that does the trick:

It turns out that the neon green diagonal contains only values in the interval 99 , 100 ] \langle 99,100] ; to the top-left everything is 99 \leq 99 ; to the bottom-right everything is > 100 > 100 . Therefore the median will lie between 99 and 100. Now focus on the values on the diagonal. The run from smallest in the corners to highest in the center. The median of these 20 values (or rather, 10 values, each of which occurs twice!) is the median of the entire table. Thus we take med { 99.1 , 99.28 , 99.44 , 99.58 , 99.7 , 99.8 , 99.88 , 99.94 , 99.98 , 100 } = 99.7 + 99.8 2 = 99.75. \text{med}\ \{99.1, 99.28, 99.44, 99.58, 99.7, 99.8, 99.88, 99.94, 99.98, 100\} = \frac{99.7 + 99.8}2 = 99.75. (The values in the spreadsheet are rounded; more accurately, we'd find 99.7475 99.7475 as the median, which Excel confirms when we type = M E D I A N ( B 2 : U 21 ) \mathtt{=MEDIAN(B2:U21)} .)

This is not exact because we picked only 400 possible length/width combinations, while there are infinitely many. Repeating this procedure with 10 000 10\,000 points, I find the median to be 99.6793 99.6793 , within 0.01% of the precise value. (Now the values between 99 and 100 no longer lie on a diagonal, but on a slightly curved line; it is part of a hyperbola.)

The solution given above applies Calculus techniques to figure out what happens if we let the number of values in the median calculation approach infinity.

Arjen Vreugdenhil - 3 years, 5 months ago

Log in to reply

Thanks for your nice graphical repesentation. :)

Paweł Samoraj - 3 years, 5 months ago
Jerry Barrington
Jan 6, 2018

In a spreadsheet, 21 rows labeled 9.025 to 10.975 in .05 steps. Same labels for 21 columns. In the grid between them, value equals row label times column label. In a separate cell, put the median function over then entire grid. Answer is 99.68, within the acceptable accuracy of the answer. (If more accuracy is needed, use more steps.)

You may have noticed, my answers tend to be more practical than theoretical.

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...