1
$\begingroup$

I have samples $x_i$ of lets say a random variable $X$ (euclidean distances, $X=\sqrt{Y}$, where $Y$ is the squared distance) which I computed from squared distances samples $y_i$. I can now calculate the sample mean $\bar{x}$ and sample variance $s^2$ as:

$$ \bar{x} = 1/n \sum_{i=1}^n x_i \\ s^2 = 1/(n-1) \sum_{i=1}^n (x_i- \bar{x})^2 $$

as described here: link

Is it possible to somehow compute the sample mean/variance for $Y$ (squared distances) (more effiecient since no square root computation needs to be done) and then transform it to the sample mean and sample variance of $X$ in an easy way?

$\endgroup$
  • $\begingroup$ The link mentions the variance of the sample mean, but not the sample variance. $\endgroup$ – drhab Jun 17 '15 at 9:39
  • $\begingroup$ jep, I corrected the question, thanks :-) $\endgroup$ – Gabriel Jun 17 '15 at 15:09
  • $\begingroup$ Link gone now. Trying to guess what is going on. See possible Answer. $\endgroup$ – BruceET Jun 17 '15 at 23:47
0
$\begingroup$

If $X_1, \dots, X_n$ are iid normal with mean $\mu$ and SD $\sigma,$ then $Z_1, \dots, Z_n,$ where $Z_i = (X_i - \mu)/\sigma,$ are iid standard normal (with zero mean and unit variance).

Finally, $Q = \sum_{i=1}^n Z_1^2$ has a chi-squared distribution with $n$ degrees of freedom, and $E(Q) = n$ and $V(Q) = 2n.$ Of course, $Q$ is the squared distance from the origin in $n$-space, if each linear component is standard normal.

If this is on the right track for you, maybe look at Wikipedia on chi-squared distribution. Also the Rayleigh distribution has to do with the (Euclidean) distance between points in a plane.

$\endgroup$
  • $\begingroup$ Where did it say the distribution is normal? $\endgroup$ – Indominus Jun 20 '15 at 5:43
  • $\begingroup$ Question unstable. Link (when present) not helpful. Context, purpose unclear. Took a stab at trying to help.See my Comment above. Notice my answer begins with "If...". If @Indominus has something useful to offer, why just a comment? $\endgroup$ – BruceET Jun 20 '15 at 17:50

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.