On Thu, 27 Feb 2003 13:45:51 +0100, David Delgado Gomez <[EMAIL PROTECTED]> wrote:
> Hello, > > I have the next problem, I would like to find a functional that the > function space is the probability distribution. > This functional should be work in this way: > Normal distribution should be 0(the minimum value) and the highest value > should be obtained for distribution that shows 2 gaussians and these > are as most separate as possible(something as dromedary's humps) You want something a test for normality, one which will score near 0 for "normal" and near its maximum for something like a dichotomy (even if it is continuous). Does that wrap it up? Many of the tests of normality behave that way, more or less. The first two tests that I think of have 0 as minimum. You won't want a test for skewness, but everything else.... I guess, specifically following your design, you could test for the empty middle -- figure out the P-levels for fractions that are *outside* of +/- 1 standard deviation. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
