With Respect
E.T.Jaynes in Probability Theory: The Logic of Science, section 7.271, pages 240-241, says that one way calling the central/Gaussian distribution "Normal" has misled statistics users is by making them think they can apply techniques such as confidence intervals that work with it to data with other distributions. Does anyone know more of the Jaynesian critique of the use of confidence intervals on nonnormal distributions than the bare reference? Does anyone know of arguments about whether that critique is valid?
I have been using confidence interval for Kendall's tau between nonnormal distributions a good deal in my work.
Yours Sincerely,
Alan E. Dunne
|