Herman Rubin ([EMAIL PROTECTED]) wrote:
: How should one decide which type of test to use EXCEPT by
: looking at its power? Statistics is not a collection of
Minor details like validity come to mind. But you're exactly right,
Herman, among tests that are valid, power is certainly an important, if
not the most important, criterion. But tests are sometimes chosen that
have a reputation for high power against corner-case alternatives over
more general tests, when these alternatives are not likely for the
context in question. Computability (both of the test statistic and its
critical values), though less often an issue, is also a relevant
criterion. Finally, interpretability and understandability are
relevant. A test or diagnostic (e.g. Q-Q or P-P plots) that gives
richer information than just a p-value may be much more valuable than a
blind test. I've seen cases where a test was originally chosen over
another because it was theoretically superior, but the superiority was
in the sixth decimal place and the method was completely unintelligible
to the intended audience (this was an application journal). Of course,
something to consider is multiple approaches: some more interpretable
and others perhaps chosen for theoretical superiority. It might be
worth pointing out that if you haven't done a histogram or Q-Q plot,
you have no business performing a test.
So yes, there are other criteria than power, but this is the first
and perhaps most important criterion to consider.
--
Clark K. Gaylord
Senior Research Engineer
Communications Network Services
Virginia Tech, Blacksburg, Virginia 24061-0506
Voice: 540/231-2347 Fax: 540/231-3928 E-mail: [EMAIL PROTECTED]
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
