suppose the sum of seven Positive number is 21. what is the minimum possible value of the average of the square of these number?
-
By numbers, you mean Whole Numbers? – hjpotter92 Mar 02 '13 at 09:39
-
And can the numbers repeat? – gnometorule Mar 02 '13 at 09:53
-
Yoni's answer below is just too intelligent for my level. I wish somebody could explain the answer to this question in simpler terms. I am not fully satisfied with Harleen's answer as well. I am wondering can't there be something less than 3 that would give us 21? like 2.5 or 1.9 or something. How immediately did we settle for 3? I wish somebody could explain the thought process to me. – aarbee Sep 12 '13 at 04:07
7 Answers
Assume $\sum_{i=1}^7 x_i = 21$. From Jensen's inequality with $\varphi(x) = x^2$ and all weights equal to $1$, we get:
$$\varphi \left(\frac {\sum_{i=1}^7 x_i} 7 \right) \leq \frac {\sum_{i=1}^7 \varphi(x_i)} 7$$
Plugging in what we know, on the left hand side we have simply $\varphi(3) = 9$, and on the right side we have the average of the squares. So the answer is at least 9.
Easy exercise: Prove it's exactly 9; that is, find 7 numbers that sum to 21 and have 9 as their average square.
- 6,817
-
It will be of great help if you can elaborate your answer in simple terms. – aarbee Sep 12 '13 at 18:07
-
@Ramit, this is a straightforward application of Jensen's inequality. I stuck to the notation in the linked Wikipedia article. What did you think should be done further? – Yoni Rozenshein Sep 13 '13 at 00:41
-
I am having a trouble understanding the terminologies of the inequality. Maybe it's beyond the scope of what I am learning. Anyways, thanks for your concern. I have understood the problem from other answers. Maybe after a few years, when I have studied enough maths, I would be able to appreciate your good answer (as indicated by upvotes.) – aarbee Sep 13 '13 at 08:03
-
$$\sum_{i=1}^n x_i = 21 \tag{1}$$
Minimizing the average of the seven sqares is the same as minimizing the sum of the squares. Both values differ by the constant facor 7. So we proceed by minizing the sum of the suqares, The latter expression is simpler so we prozeed by minimizing
$$\sum_{i=1}^7x_i^2 \tag{2}$$.
We can write
$$ \begin{eqnarray} \tag{3} \\ x_1=3+d_1 \\ x_2=3+d_2 \\ \cdots \\ x_7=3+d_7 \\ \end{eqnarray} $$
Substituting $(3)$ in $(1)$ gives
$$\sum_{i=1}^n d_i = 0 \tag{4}$$
So we have
$$ \begin{eqnarray} \sum_{i=1}^7 x_i^2 = \\ \sum_{i=1}^7 (3+d_i)^2 = \\ \sum_{i=1}^7 (9 + 6 d_i +d_i^2) = \\ 63 + 6 \sum_{i=1}^7d_i + \sum_{i=1}^7 d_i^2 = \\ 63 + \sum_{i=1}^7 d_i^2 \tag{5} \end{eqnarray} $$ The sum of $d_i$ vanishes because of $(4)$ $(5)$ is $\gt 63$ if some $d_i \gt 0$. It is equal to 63 and therfore minimal if all $d_i=0$.
This is equivalent to all $x_i=3$.
- 11,049
-
1Thanks for such a such a sweet answer. I understood it completely. No doubt left. I will award the bounty soon. – aarbee Sep 13 '13 at 08:04
Let's assume we have a 7-tuple in which some numbers are not equal. Call two numbers that are different $x$ and $y$, then $\frac {x^2+y^2} 2$ is more than $(\frac {x+y} 2 )^2$, which is the average if $x$ and $y$ were each replaced by their average, $(\frac {x+y} 2)$.To show this, first expand $(\frac {x+y} 2 )^2$ to $\frac {x^2+y^2+2xy} 4$. Then, to show that $\frac {x^2+y^2} 2>\frac {x^2+y^2+2xy} 4$ multiply by $4$, you get $2x^2+2y^2>x^2+y^2+2xy$. Subtract $x^2+y^2+2xy$ from both sides and get $x^2+y^2-2xy>0$, which is clearly true if $x \ne y$ because it's equal to $(x-y)^2$. This shows that any such "reduction" will always lower the average of squares.
Now, assume that there is a 7-tuple with a lower average than $(3,..,3)$. You can easily reduce it with my method to a 7-tuple with four numbers equal, and the other three equal as well.(i.e. $x,x,x,x,y,y,y$.) Just apply the reduction method to the first two numbers, the 3rd and 4th,1st and 3rd,2nd and 4th,5th and 6th,4th and 7th,4th and 5th,and 6th and 7th, in that order.(Try it with an actual group of distinct numbers if you don't follow:)
Now this 7-tuple's average of square is lower than our original 7-tuple, and so that average must be lower than 9. However this cannot be for the following reason. The sum of the squares in terms of $x$ is: $4x^2+3((\frac {21-4x} 3)^2)$. This expands to $\frac 7 3 (4x^2-24 x+63)$. For this to be less than 63, $7(4x^2-24 x+63)$=$28x^2-168x+441$ must be less than 189, so $28x^2-168x+252=28(x-3)^2$ must be less than zero. This clearly cannot be for any real x.
Therefore, any solution that all the numbers are not the same cannot have an average square of less than that of $(3,..,3)$, namely 9.
QED
- 170
-
-
@miracle This was supposed to show that no other answer can be the minimum, but I had a stupid typo which I now fixed. Given any other answer, I construct one with a lower average. – ike Sep 15 '13 at 19:34
-
You have a method that for a tuple $(x_1,\ldots,x_7) \ne (3,\ldots,3)$ constructs a different tuple $(y_1,\ldots,y_7)$ with a smaller squaresum. For $(3,\ldots,3)$ the method does not work because the constructed tuple is $(3,\ldots,3)$ again. From this you conclude the $(3,\ldots,3)$ has the smallest squaresum. That conclusion is not valid. – miracle173 Sep 15 '13 at 21:11
-
@miracle Why not? Assuming there is a number group that has the lowest "squaresum", it must be (3,...,3) or it's not the lowest. – ike Sep 16 '13 at 17:25
-
yes, if you assume there is a tuple that has the lowest squaresum. but you nowhere proved that there exists such a tuple. – miracle173 Sep 16 '13 at 18:44
-
Let's answer a more general problem: Given numbers $x_1, x_2, \dots,x_n$ such that $$ \sum x_i=\sum_{i=1}^n=S $$ we want to find values $x_i$ such that $$ \sum x_i^{\,2} $$ is minimized.
For reasons that will be clear in a moment, consider the sum $\sum(x_i-(S/n))^2$. We'll have $$\begin{align} \frac{1}{n}\sum\left(x_i-\frac{S}{n}\right)^2 &= \frac{1}{n}\sum\left(x_i^{\,2}-2\frac{S}{n}x_i+\left(\frac{S}{n}\right)^2\right)\\ &=\frac{1}{n}\sum x_i^{\,2}-\frac{2}{n}\frac{S}{n}\sum x_i+\frac{1}{n}\sum\left(\frac{S}{n}\right)^2\\ &=\frac{1}{n}\sum x_i^{\,2}-2\frac{S}{n}\frac{S}{n}+\frac{1}{n}n\left(\frac{S}{n}\right)^2\\ &=\frac{1}{n}\sum x_i^{\,2}-\left(\frac{S}{n}\right)^2 \end{align}$$ Consequently, we have $$ \frac{1}{n}\sum x_i^{\,2}=\frac{1}{n}\sum\left(x_i-\frac{S}{n}\right)^2+\left(\frac{S}{n}\right)^2 $$ In other words, the left-hand side, i.e., the average of the squares, will be minimized when the right-hand side is minimized. That clearly will happen when all of the terms $x_i-(S/n)$ are zero, which can only happen when all the $x_i=S/n$, which is to say when each $x_i$ is equal to the average of all the $x$s.
In the original problem, with $n=7$ and $S=21$, that will happen when each $x_i=3$.
- 8,718
we need 7 numbers whose squares's average should be minimum, i.e. we need smallest of numbers to be added.
Simply we can observe
21=3+3+3+3+3+3+3
thus average of squares of these 7 numbers = (9+9+9+9+9+9+9)/7=9
Consider the formula $V(X)=E[X^2]-(E[X])^2$ for variance, which is necessarily non-negative.
- 79,832
$$ {\cal F} \equiv {1 \over 7}\sum_{n = 1}^{7}x_{n}^{2} - \mu\left(\sum_{n = 1}^{7}x_{n} - 21\right) $$ $$ 0 = {\partial{\cal F} \over \partial x_{n}} = {2 \over 7}\,x_{n} - \mu \quad\Longrightarrow\quad x_{n} = {7 \over 2}\,\mu \quad\Longrightarrow\quad \mu = {42 \over 49} \quad\Longrightarrow\quad x_{n} = 3 $$
$$ {\cal F} = {1 \over 7}\sum_{n = 1}^{7}\left(3 + \epsilon_{n}\right)^{2} = 9 + {6 \over 7}\overbrace{\sum_{1}^{n}\epsilon_{n}}^{0} + {1 \over 7}\sum_{n = 1}^{7}\epsilon_{n}^{2}\quad \color{#ff0000}{\large\geq\ 9} $$
- 89,464