I would appreciate it if someone could clear-up for me what the "power
of a statistical test" really means.
I have been looking at the excellent book "Goodness-of-Fit Techniques"
by D'Agostino and Stephens where
power is an important topic.
My understanding was that the power function must be between 0 and 1;
but, in this book results for "Power Studies" are not in
this range. Example, Table 10.6 on pp. 452-453 of this book gives power
studies results (generated by Stephens). It is stated on p. 451 that
" ... has conducted a large power study on the various procedures for
H_0, using Monte Carlo samples..."
I thought a power analysis was about H_1 (the alternative hypothesis).
How were these entries in Table 10.6 actually generated?
I seem to missing something important in their use of "power".
--Thanks,
V. Stokes
|