To summarise, the question is not

(p-value:) "Can we tell that this data apart from normal?"  (since we already
know it isn't)

nor really even

(test statistic:) "How much does this deviate from normality?"

but

"How much is that non-normality going to hurt what we're doing?"


However, in certain restricted circumstances, you might be able to use a good
test-statistic as /part/ of your understanding of the answer the the third
question...

Glen

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to