[EMAIL PROTECTED] (Paul R Swank) wrote in 007d01c290b4$adee3de0$520e6a81@PEDUCT225:">news:007d01c290b4$adee3de0$520e6a81@PEDUCT225: >the websites posted below gave me some problems Cannot explain why it couldn't be found. It's always been there when I looked. The second link required either that I get a newsreader that doesn't wrap (sorry, I like my newsreader) or that you cut and paste the broken URL. It's still there when I cut and paste from your reply and delivering graphical depictions of non-central t's. > If the > assumptions for t are met, that is, the observations are NID with equal > variances, why would the non-central t be asymmetric. > Think of the joint probability of the normal pdf on the x-axis and the chi- square pdf with more than one df on the y-axis. In the central-T you are "looking" at the joint distribution from "below" the middle of the normal distribution (x=0), whereas in the non-central T you are "looking" (double integration, really) at the joint distribution from the origin. The joint density is shifted over to the right or left depending on the non- centrality parameter. The asymmetry comes from the side-long look at the chi-square (or chi) distribution along lines radiating out from the origin with various slopes.
Let me admit that if I were using my "intuition" that I would have predicted that the skewness would be opposite to what happens, but that probably comes from my not being able to adjust for the ratio of x/y in the integration. David Winsemius. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
