- Joined
- Jun 21, 2003
By the way, the "trimmed mean" is a very troublesome statistic, verging over into the weed-infested field of "non-parametric staistics." Except in cases where certain smoothing conditions obtain (notably symmetry), the trimmed mean does not follow any well understood distribution of values, and various "bootstrap" methods are often called upon.
If you try to talk about standard deviations, for instance, you have to use the full n=9 for sample size, even though only 7 scores contribute to the "mean."
If you try to talk about standard deviations, for instance, you have to use the full n=9 for sample size, even though only 7 scores contribute to the "mean."
)
