5
$\begingroup$

I have a list of (x,y) pairs that form a cluster of points. I was able to find the mean of the cluster, but I'd like to be able to find the standard deviation of the cluster. How can I define the standard deviation? I defined the mean as the point (mean of x values, mean of y values).

What type of standard deviation value is typical for clusters like this? I'd imagine I could get a standard deviation of the spread, or something like that?

This looks good. Gaussian on a cluster:

Gaussian Data

How do I do this?

Improve this question
$\endgroup$
2
  • 1
    $\begingroup$ Obtain within-cluster covariance matrix between X and Y, sum the two diagonal elements and take the root. $\endgroup$
    – ttnphns
    Commented Apr 6, 2012 at 7:14
  • $\begingroup$ Two comments: 1. Great graphic. It looks like you would have to have something like an implicit standard deviation in the clustering process, or else how would you judge points like 0.8,0.9 to belong to the red cluster, not the blue one? $\endgroup$
    – Owe Jessen
    Commented Apr 6, 2012 at 13:04

1 Answer 1

Reset to default
3
$\begingroup$

The two easiest approaches:

  1. One way to do it would be to count the standard deviation on $x$s and $y$s separately, to get you a box shape.

  2. If you prefer the circle shape, then simply given the mean $(\bar{x},\bar{y})$ and the points $(x_i,y_i)$ compute the distance $r_i = \sqrt{(\bar{x} - x_i)^2 + (\bar{y} - y_i)^2}$ and then calculate the standard deviation of the $r_i$'s around $0$. The resulting standard deviation (or if you divide by $\sqrt{N}$ then standard error) will be the radius of the circle.

Improve this answer
$\endgroup$
2
  • $\begingroup$ "then calculate the standard deviation of the ri's around 0", could you please explain what you mean by std around 0? $\endgroup$
    – Ankit Agrawal
    Commented Jul 26, 2019 at 11:48
  • $\begingroup$ if you work with 2d array of vectors (xi, yi), you can count std per each axis - then it'll be a box, but if you need radius instead of a box you can calculate distances between each vector and mean of a set. You'll get bunch of 1d values of distances spreaded around 0 since we already subtracted mean on previous step. Now you just need to calculate std of distances with mean=0 $\endgroup$
    – Sergey Skrebnev
    Commented Jun 30, 2021 at 6:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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