Joe wrote: > > Couldn't you read my mind? Sorry for the confusion but what I should have > written is: > > (3*mean - median)/std. dev.
Still looks wrong. For example, now left-skewed distributions will have a positive skewness by this measure. I'm sure that an error crept in somewhere between the original expression of this and when it got to you. In fact I bet that the original from which this came was just the 3*(mean - median)/std. dev. you had before. > So thanks for the discussion, but it sounds like I could create a skewness > formula sqrt(mean^2-median^2)/stnd. dev - call it Joe's formula for > skewness, and get it published somewhere - even if its only on google. You could certainly create a new formula - after all, thats where new formulas come from - people come up with them for one reason or another. If you wanted it published somewhere anyone would take notice, you'd at least need to explain why it was more useful in some circumstances than the already existing measures. Glen . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
