PDA

View Full Version : Re: Statistical Power Analysis



slawson74
01-22-2005, 07:32 AM
Dear Alistair,

Presuming that you are using power analysis to
calculate a suitable sample size for a given effect
size, variation (SD), alpha level (probability of a
Type I error), and power value (Power to detect an
effect, 1 - b, where b is the probability of a Type II
error):

The popular consensus generally seems to be the rather
arbitrary use of alpha =.05 and Power=0.80. (Perhaps
someone can explain this consensus to me.) In my
opinion there is no such thing as a correct cut off
value so don't beat yourself up over it too much.

Knowing the size of the effect that you want to detect
is more difficult. Without knowing what your subjects
would be doing, or what intervention you are making,
it is hard to comment on the effect or the variation
that you are likely to see. The most common solution
to this problem is to conduct a short pilot study with
limited numbers to give you an idea of the size of the
changes you'll witness and the general variation in
the response variable.

If you are looking at lower limb kinematic and
kinetics during human walking, an excellent source of
normative values and variations for is Chris Kirtley's
Clinical Gait Analysis site:
http://guardian.curtin.edu.au/cga/data

Best of luck,
Sian

Dr Sian E M Jenkins Lawson
Biomécanique et Pathologie Locomotrice de Cheval
Ecole Nationale Vétérinaire d'Alfort
7, Avenue du Général de Gaulle
94704 Maisons-Alfort cedex FRANCE
Tel : 01 43 96 72 49
Fax : 01 43 96 31 62
Email: sian@equinemechanics.com
Web: http://www.equinemechanics.com

-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
-----------------------------------------------------------------