|
Greg Kochanski | |
(In statistics, that is...)
If you have (or if you assume you have a Gaussian distribution), then a z-score of 0 corresponds to the middle of the distribution, a z-score of 1 corresponds to an event that is 1 standard deviation above the mean, and z-scores bigger than 3 or smaller than -3 correspond to rare events in the tails of the distribution.
If you have measurements like positions, the z-score is a simple linear scaling of the data. If you have measured probabilities, like "the fraction of the class that failed the test", then the z-score is the inverse of the cumulative distribution function for a Gaussian probability distribution. That's an s-shaped nonlinear function. For more details, see [ 1, 2, 3, or 4 ] for more details.| [ Papers | kochanski.org | Phonetics Lab | Oxford ] | Last Modified Thu Sep 1 10:55:27 2005 | Greg Kochanski: [ Home ] |