"Phillip Good" <[EMAIL PROTECTED]> wrote in message news:[EMAIL PROTECTED] >Rejecting a recent article of mine on the behavior of ANOV with respect to non-normal errors and >small samples, an editor said I'd failed to cite the many articles already on that subject. A >literature search back to 1960, elicited exactly one article on point, though I found many >concerning the effects of unequal variance or dependence. Can you suggest a couple of articles?
>Phillip Good I am not clear quite what you are looking for, but a quick look in a few issues found .. Jensen D.R (1979) Linear models without moments. Biometrika 66 (3) 611-7 ... shows good properties of estimators for non-normal residuals, even if moments don't exist. Davis A.W. (1980) On the effects of moderate multivariate nonnormality on Wilks's likelihood ratio criterion. Biometrika 67 (2) 419-27. ... this may be rather more general than you wanted, being multivariate ANOV. But it contains the following reference... Gayen A.K. (1950) The distribution of the variance ratio in random samples of any size drawn from non-normal universes. Biometrika 37, 236-55. You might also try looking for references in some standard deep text books ... presumably they comment on this topic at least. David Jones . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
