6
$\begingroup$

I know that $\bar{x}$ makes absolute result of $\sum_{i=0}^n (x_i-\alpha)$ minimum. In fact it makes it zero. But how to find what value of $\alpha$ makes $\sum_{i=0}^n (x_i-\alpha)^2$ minimal? What is the best approach?

$\endgroup$
  • 2
    $\begingroup$ By "the absolute result of $\sum (x_i-\alpha)$", do you mean $|\sum (x_i-\alpha)|$? $\endgroup$ – Glen_b Oct 11 '15 at 9:23
  • 1
    $\begingroup$ Since $\sum (x_i-\alpha)= 0$ when $\alpha=\bar{x}$, you're correct that this value for $\alpha$ minimizes $|\sum (x_i-\alpha)|$. As to the question of what value for $\alpha$ minimizes $\sum (x_i-\alpha)^2$, this can be achieved in several ways, such as by simple calculus, or by writing $\sum (x_i-\alpha)^2=\sum (x_i-\bar{x}+\bar{x}-\alpha)^2$ and then expanding and simplifying. I think the first approach is covered in a number of questions already on our site; the second approach may be as well. I'll see if I can find some links $\endgroup$ – Glen_b Oct 11 '15 at 9:39
  • $\begingroup$ I found some posts that related to those two solutions, but nothing quite close enough to really call a duplicate. I'll outline one approach. $\endgroup$ – Glen_b Oct 11 '15 at 11:05
  • 3
    $\begingroup$ As an interesting aside, the median minimizes the sum of the absolute deviations:$|x_i-\alpha|$ $\endgroup$ – user75138 Oct 11 '15 at 12:26
  • 2
    $\begingroup$ As stated in this answer, It is well-known (or easily proved) that the quadratic $\alpha z^2 + 2\beta z + \gamma$ has an extremum at $z = -\frac{\beta}{\alpha}$. This shows that, for any given $n$ real numbers $x_1, x_2, \ldots, x_n$, the quantity $$G(a) = \sum_{i=1}^n (x_i-a)^2 = \left(\sum_{i=1}^n x_i^2\right)-2a\left(\sum_{i=1}^n x_i\right) + na^2,$$ has minimum value when $\displaystyle a = \frac 1n \sum_{i=1}^n x_i =\bar{x}$. $\endgroup$ – Dilip Sarwate Oct 11 '15 at 14:32
13
$\begingroup$

As I mentioned in comments, showing what minimizes $\sum (x_i-\alpha)^2$ can be done in several ways, such as by simple calculus, or by writing $\sum (x_i-\alpha)^2=\sum (x_i-\bar{x}+\bar{x}-\alpha)^2$. Let's look at the second one:

$\sum (x_i-\alpha)^2=\sum (x_i-\bar{x}+\bar{x}-\alpha)^2$ $\hspace{2.55cm}=\sum (x_i-\bar{x})^2+\sum(\bar{x}-\alpha)^2+2\sum(x_i-\bar{x})(\bar{x}-\alpha)$ $\hspace{2.55cm}=\sum (x_i-\bar{x})^2+\sum(\bar{x}-\alpha)^2+2(\bar{x}-\alpha)\sum(x_i-\bar{x})$ $\hspace{2.55cm}=\sum (x_i-\bar{x})^2+\sum(\bar{x}-\alpha)^2+2(\bar{x}-\alpha)\cdot 0$ $\hspace{2.55cm}=\sum (x_i-\bar{x})^2+\sum(\bar{x}-\alpha)^2$

Now the first term is unaltered by the choice of $\alpha$ and the last term can be made zero by setting $\alpha=\bar{x}$; any other choice leads to a larger value of the second term. Hence that expression is minimized by setting $\alpha=\bar{x}$.

$\endgroup$
5
$\begingroup$

Putting the first derivative (with respect to $\alpha$) equal to zero you find $2\sum_i (x_i -\alpha) (-1) = 0$ so $\sum_i x_i = n \alpha$ or $\alpha = \frac{1}{n} \sum_i x_i$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.