we have to first separate out 2 things:
1. some test statistics are naturally (the way they work anyway) ONE sided
with respect to retain/reject decisions
example: chi square test for independence ... we reject ONLY when chi
square is LARGER than some CV ... to put a CV at the lower end of the
relevant chi square distribution makes no sense
2. whether for our research hypothesis ... rejection of the null is
something that makes sense to BE ABLE to do regardless if the evidence
suggests that the effect is LESS than the null or MORE than the null
example: typical treatments could have positive or negative effects (even
though obviously, we predict + effects) ... thus, when doing a typical two
sample t test (if you are interested in differences in means) ... we make
both an upper AND lower rejection region ... ie, two tailed TEST
but, in some cases, it might be totally unthinkable for one end of the
statistical distribution to be "useful" in a given case ... say we have a
weight loss regimen program ... consisting of diet and exercise ... and
want to know if it works ... ie, people lose weight ... now, in this case
(it could be) one might argue that it is difficult to conceptualize that
the regimen would actually "cause" one to GAIN weight ... so, to put some
rejection area on that end of the t distribution would seem silly ... thus,
we might be able to make the case that it is perfectly legitimate to use a
one tailed test in this case ... (done BEFORE hand of course ... not just
after the fact because your 2 tailing approach failed to allow you to
reject the null)
At 03:08 PM 3/13/01 +1300, Will Hopkins wrote:
>At 7:34 PM +0000 12/3/01, Jerry Dallal wrote:
>>Don't do one-tailed tests.
>
>If you are going to do any tests, it makes more sense to one-tailed
>tests. The resulting p value actually means something that folks can
>understand: it's the probability the true value of the effect is opposite
>to what you have observed.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
