PDA

View Full Version : Single step analysis of Ground Reaction Forces During Gait



Dewitt, John K. (jsc-sk) (wls)
11-16-2003, 10:27 AM
Hello All,
A while back I posted a question to the list regarding the treatment of
steps during gait as individual measurements. The following is the original
question and the responses I received:

Original Post:
Hello All,
I am looking for some advice and suggestions on this biomechanics data
analysis issue.

I have GRF data from 3 subjects during locomotion under varying conditions.
I was only able to collect 1-2 trials at each condition, but I do have
multiple footfalls per trial. I would like to compare certain components of
the GRF curve (i.e. 1st peak, 2nd peak, impulse) between conditions. If I
were to find an average for each of these variables for each trial, I end up
with 1-2 measures per person, which will not allow me to do statistics. I
would like to know if it is acceptable, given that this is a repeated
measures design where each subject will be compared to themselves, to take
multiple (i.e. 5-10) footfalls and consider each to be a trial. In this
approach, the number of trials now becomes 5-20 per person.
1) Is this acceptable?
2) Are there any suggested references that may have used this
approach in their design?

Thanks, and as usual I will summarize the responses
=============================================
Responses:
John,

You should be able to go over a couple of buildings from where you are and
pick up these papers from Jacob Bloomberg or Ajit Mulavara. They may be of
some value to you. Jacob also published a "loco" paper that Vernon and I
are not authors on but was from the same sort of data set. The low n issue
is always interesting. Good luck.

Dr. Layne

Layne, C.S., Lange, G.W., Pruett, C.J., McDonald, P.V., Merkle, L.A., Smith,
S.L., Kozlovskaya, I.B., Bloomberg, J.J. Adaptation of Neuromuscular
Activation Patterns During Treadmill Walking After Long-Duration Space
Flight. Acta Astronautica 43, 107-120, 1998.

Layne, C.S., McDonald, P.V., & Bloomberg, J.J. Neuromuscular Activation
Patterns During Treadmill Walking after Space Flight. Experimental Brain
Research. 113(1) 104-116, 1997.


McDonald, P.V., Basdogan, C., Bloomberg, J.J., Layne, C.S. Lower Limb
Kinematics During Treadmill Walking After Space Flight: Implications for
Gaze Stabilization. Experimental Brain Research 112 (2) 325, 1996.
============================================

John,
The way that you describe it you want to do single subject analysis. Dr
Barry T Bates and his colleagues have done a lot of work in this area
regarding between and within subjects variability. So, just do a search
under his name in PUBMED. Also, he recently wrote a nice chapter in the
attached book.
Nick
NOTE - The attachment was a flier for the book: Innovative Analyses of Human
Movement
=======================================
Hi John,
Yes there are acceptable and appropriate statistics for you problem.
Try a mixed effects model (Pinheiro, JC and Bates, DM. Mixed-effects models

in S and S-plus. New York, Springer-Verlag. 2000 This book is nearly
incomprehensible to us non-statisticians so good luck reading it). The good

news is that these mixed effects models allow you to model trials as a
random effect within a repeated measures ANOVA. If you take the mean of
each subject not only will you reduce your statistical power, you will also
hide some of the true variability. In all of statistics its the ratio of n
(number of subjects) to variance that is the most important. From a
validity and statistical power standpoint you always win (have a greater
chance of detecting significance) when you count trials as subjects, even if

the variability goes up somewhat. This is inappropriate in most ANOVA
models which consider each measure as coming from an individual rather than
each individual contributing multiple measures. Repeated measures ANOVAs
account for the decrease in population variability due to each subject
serving as their own control, but they do it at a very high price in terms
of statistical power reduction. The mixed effects models account for this
reduction in variability due to a number of samples coming from each subject

without such a steep price in statistical power reduction. They have been
around for some time, and are perfectly suited to the real world of limited
time, limited grant dollars and limited subjects.

Michael
===========================================
John

It looks to me that you have a 2 way repeated measures design with
CONDITIONS and TRIALS as your treatment variables. An ANOVA analysis using
this design will allow you to include all trials from all subjects in your
analysis. The TRIAL main effect will provide you additional insight as to
any changes in your experimental variables across trials for each subject
and your COND x TRIALS interaction term will provide insights to any changes
across trials as a function of conditions. I use this experimental design
in work in my lab. In my opinion, this is a straightforward experimental
design issue and it can be explained in the methodology of a manuscript. Any
decent stats text or experimental design text should serve as a valid
reference.

Hope this helps,

Michael

________________________
Michael E. Feltner, Ph.D, FACSM
Dept. of Sports Medicine
Pepperdine University
Malibu, CA 90263 USA
EMAIL: michael.feltner@pepperdine.edu
WEB: http://faculty.pepperdine.edu/mfeltner/

VOICE: (310) 506-4312
FAX: (310) 506-4785
===============================================
Thanks to all that responded




John DeWitt, M.S., C.S.C.S.
Biomechanist - Exercise Physiology Laboratory
Space Physiology & Countermeasures
NASA - Johnson Space Center
Houston, TX 77058
281-483-8939 / 281-483-4181 (fax)


-----------------------------------------------------------------
To unsubscribe send SIGNOFF BIOMCH-L to LISTSERV@nic.surfnet.nl
For information and archives: http://isb.ri.ccf.org/biomch-l
Web-based discussion forum: http://movement-analysis.com/biomch_l
-----------------------------------------------------------------