On 19 Jan 2003 18:32:09 -0800, [EMAIL PROTECTED] (Glen) wrote:

[ snip, answers to the questions.]
> 
> ** if both give similar p-values there is no issue. If they
> don't (e.g. one has p=.3 and the other p=.01), it may be an 
> indication that an assumption is being violated for one of 
> them - in which case you need to be cautious about interpreting 
> the results until you understand why they differ so much.
> 

Good advice.  I wanted to repeat it, and add on.

People ask, What does it matter if you violate assumptions?
It  *is*  possible to say that "a Pearson correlation is still a 
Pearson correlation,"  because that is meaningful 
beyond its possible use as a test.  But it might not be very
meaningful to compare it to other Pearson r's.

The bad effect of "violating assumptions", if it doesn't
destroy the meaning of the measure, is that  it can change 
the *test* .  You might get a different p-value, than for 
another test of what-seems-to-be the same hypothesis.  

If you get the same p-value when you test in all the possible ways, 
then you apparently aren't doing anything 
that offends the assumptions in any *notable*   way.  
On the other hand, if the p-values vary, then you 
have a problem until you figure out why.

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
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