On 3 Nov 2000 15:45:52 -0800, [EMAIL PROTECTED] (Paul Bernhardt)
wrote:
< snip, various stuff on the problem >
> But, I'd be uncomfortable with data with only 2 observations in one
> group. If you are stuck with the data, then you are stuck with the equal
> variances t-test because there is insufficient evidence on which to base
> a decision to use the unequal variance t-test.
>
- When I was reading up on the problem,
I liked the comment that I ran into in one paper,
"No one who has studied the problem ever
recommended the practice of selecting 'which t-test'
on the basis of a test of variance."
I liked that comment, even though those authors
ended up preferring non-pooled t tests as a general tool.
As I said in another note, if you have unbounded
scores, you should see if transformation fixes the
problem. (For scale scores, just don't worry.)
After that, I think I like to see consistent outcomes
for various tests, including non-pooled, Randomization,
and a test on ranks.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
