Thanks Tim. You wrote: "If that's all the information you have, you can't decide for either Y1 or Y2 whether it is "fittable" by a gaussian distribution. That's a value judgement which depends upon the purpose and tolerances on "fitting" " You are right. In fact this is true in any cicustance. About the same problem, I would foucs your attention on my new post: Probability Plot Correlation Coefficient Plot and Tukey lambda http://groups.google.com/group/sci.math/browse_thread/thread/cacbe0a0e1deb6ce#
Do you think this approach is better? Overall, you would be able to show me how this procedure has to be implemented?