February, 2003
© 2000, 2001, 2003 Scott S. Emerson, M.D., Ph.D.
In the
presence of sampling error, we must be careful in how we define
hypotheses
In the
presence of sampling error, an infinite sample size is required to be
able to address both of those contiguous hypotheses with high
precision simultaneously
Possible
solutions
·Treat
hypotheses asymmetrically
·Define
experiment in terms of noncontiguous hypotheses