On Fri, 28 Feb 2003 17:21:45 -0500, Rich Ulrich <[EMAIL PROTECTED]> wrote: > On 28 Feb 2003 12:18:18 -0800, [EMAIL PROTECTED] (saisat) > wrote: > >> Hi Rich, >> >> I went through ANOVA. But as far as i can understand it only tells me >> if the different distributions are similar or not. It does not tell me >> which ones to remove. is there any other techniques to do that >> > referring to ME? > > I said, People who know statistics don't ask that question. > It is usually improper (WRONG) to do. > > By the time you learn enough to know that it is > something REASONABLE -- by learning the > statistics, and decision theory, etc. -- it is almost sure that > you would have come across your criterion along the way. > > If it still seems important, consult with a statistician in person.
Perhaps I am excessively naive here and need more eludication. What about testing the composite hypotheses of "equal means" consisting of all possible poolings of discrete sets? Which is best justified by the data against alternatives? How would one do that? -- --------------------------------------------- Matthew Kennel, [EMAIL PROTECTED] Institute For Nonlinear Science, UC San Diego --------------------------------------------- . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
