Don Burrill in a discussion of the assumptions of t-test assumptions writes:
> "Acceptable" I don't know about: depends on the universe of discourse.
> But you should try to justify the assumption that the observations are
> taken independently, and that the underlying within-group variances are
> approximately equal. And you should also be aware that while the t-test
> is well known to be fairly robust against violations of assumptions,
> that robustness applies to two-sided tests; one-sided tests are, by
> comparison, rather fragile.
>
most of the robustness studies pertain to Type I error rates. With
non-normal distributions, the p-values for Student's t for the null
hypothesis are pretty close. Such is often not the case for Type II error
rates. Long-tailed distributions can just about eliminate any possibility
of finding group differences. Use normal quantile-quanitle plots to check
for non-normality, especially in the tails. If it is big enough to cause
problems, you will be able to see it in the plot.
gary
--
[EMAIL PROTECTED] http://psych.colorado.edu/~mcclella/
Dept of Psych, CB 345, Univ of Colorado, Boulder, CO 80309-0345 USA
voice: 303-492-8617 fax: 303-492-5580
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
