View Full Version : summary of responses

02-28-2007, 02:15 AM

Thank you all for the replies concerning my question.
Maarten Oosterlinck

Original question:
I have a question about the 'normalization' of GRF data. I read that there is no linear correlation between GFR and weight. How should GRF then be compared between individuals? Is the normalization to weight no longer done? I read that some people even normalize GRF data to weight and to stance time. Which normalization is generally accepted?

I believe linear normalizing GRF by body weight is very common, and effective in accounting for differences in subjects' body masses. However, when normalizing loading rates, body weight may not be the most appropriate normalization variable. There was a paper published on this topic recently that you might find useful:

Mullineaux DR, Milner CE, Davis IS, and Hamill J (2006). Normalization of ground reaction forces. Journal of Applied Biomechanics, 22, 230-233.

Ross Miller
6 Totman Building
University of Massachusetts Amherst

(Remark: in the paper mentioned above, normalisation is done to bodymass, instead of to weight as in a lot of other papers. Maarten )

A nice paper on normalization of biomechanical data is G. Markovic and S.
Jaric (2004) Eur J Appl Physiol 92:139-149.
Not sure who wrote that "there is no linear correlation between GRF and
weight": there is certainly a correlation (equality) between GRF and weight
during quiet standing...
How and whether to normalize depends on the questions you are asking.
Bill Rose
University of Delaware, USA

Where did you read this? Do you mean there is no correlation between
subjects of different weights walking at the same velocity.
Is this velocity measured as the CoM velocity or the step or foot velocity.
As you might appreciate the step/foot or stride velocity would be slower for
a tall man with the same CoM velocity as a short man.
This is why it may be more useful to normalise GRF to weight and limb length
or limb velocity. Most paper I have read, that normalise, usually normalise
to body weight only. This can be misleading for the above reason.

Is this helpful All the best Dave Smith

Many thanks to all of you who responded to my query regarding normalisation
of forces and moments to bodyweight/mass and height.

>From your responses it has become clear that normalisation means
representing the data as a unitless value which enables a more acuarte
comparison between large groups of subjects who may vary widely in their
stature. So, a Force (in Newtons) would be normalised to Body weight (also
in Newtons) and that a joint torque/ Moment (in Newton-metres) would be
normalised to bodyweight (Newtons) * height (metres). Height could represent
the overall height of the subject or as some of you have recommended, the
limb length of each subject.

Armed with this new information I re-read some of the journal articles and
found that a few authors have written bodymass in the text but have actually
normalised to body weight- hence my original confusion!! I hope other
people find the information below as useful as I did. I have posted my
original query followed by a summary of responses.

Thanks again,

Susan Wilson

It makes sense to normalise vertical GRF data by dividing by body weight
(in N) because, from basic mechanics, we know that mean vertical GRF
must be equal to body weight. However, this method will not remove all
size effects if your subjects are running at the same speed because, in
this case, the smaller subjects are running at a higher speed relative
to their size. In order to remove the effects of size you would need to
have your subjects running at the same relative speed, rather than the
same absolute speed. Using equal values of the Froude number
(speed/sqrt(gravitational acceleration*leg length)) would probably work,
but it's not guaranteed to (this will depend upon whether your subjects
are moving in a dynamically similar manner).

Horizontal GRFs should not be divided by body weight because the above
argument about mean GRF does not apply. You could divide by mass, which
would give the horizontal acceleration of the centre of mass (assuming
other horizontal forces are insignificant).

I don't see any good reason for dividing vertical GRF by body weight and
stance time. If the vertical GRF trace is sinusoidal (which it is
approximately), then the following equation applies: F/mg =
C*t_stride/t_stance, where C is a constant t_stance = stance time,
t_stride = stride time, mg = body weight and F = peak vertical GRF
(Alexander et al., 1979, J Zool 189: 135). If we divide through by
stance time we get: F/(t_stance*mg) = C*t_stride/(t_stance^2). I don't
know of any reason why the right hand side of this equation should be
independent of the size of the subject so I don't think this is a good

In summary, normalising GRF data to remove size effects is not
straightforward. If possible, use subjects that are approximately the
same size. If this is not possible, I suggest dividing vertical GRF by
body weight and horizontal GRF by mass and comparing subjects moving at
speeds corresponding to equal Froude number. However, it would still be
a good idea to use subjects of a wide range of different sizes and to
check whether this approach has been successful in removing size effects
by testing statistically whether there is a significant relationship
between your measured parameters and body size (mass or leg length).

Feel free to get in touch if you would like to discuss this further.

Good luck with your work,

Sharon Bullimore

Where have you read that?
In all equations of movement body mass and gravity play a role, so normalizing GRF to body weight is perfectly sensible.
An other point is that GRF is sensitive to speed, and that this speed should also be normalized to make a comparison possible.

Hartelijke Groeten,
At Hof
Centrum voor Bewegingswetenschappen
Rijksuniversiteit Groningen
Postbus 196
9700 AD Groningen

I think you need to express GRF vertical axes as %BW (body weight) and
horizontal axes as %cycle. This will allow comparison, inter and intra

Hamid Rassoulian