4 ms·
The author makes a minor mistake. You can make an unbiased estimate of the variance using the formula s^2 = (sum of squared deviations) / (N-1). However, takin
by yummyfajitas 16y ago
The author makes a minor mistake.
You can make an unbiased estimate of the variance using the formula s^2 = (sum of squared deviations) / (N-1). However, taking the square root of this gives you a biased estimate of the standard deviation due to the concavity of sqrt.
http://en.wikipedia.org/wiki/Bessels_correction http://en.wikipedia.org/wiki/Bessels_correction
- wnoise 16y agoHeh. Of course being a non-linear transformation, bias will creep in if you just naively take the non-biased value, but people rarely think about that. Thanks for pointing this out. (BTW, the link needs an apostrophe: http://en.wikipedia.org/wiki/Bessel%27s_correction http://en.wikipedia.org/wiki/Bessel%27s_correction )