velocity normalization

[Petcu Daniel]

Dear Members,

I'm facing with the next problem, asking your help : I would like to compare the effect of an intervention on the free moment. As the values of the free moment are influenced by the subject velocity and possibly, by the intervention I would like to reduce the velocity's contribution through normalization. The only one reference regarding the velocity normalization I found in the case of angular momentum [ http://jeb.biologists.org/content/211/4/467.abstract ]. As there are big differences between subject velocities pre and post intervention I would like to ask you if free moment normalization with velocity x body weight x height could solve my problem. I would be grateful for any references on this subject.

Thank you in advance for your help,

Respectfully,
Daniel

[Ross Miller]

Re: velocity normalization

Hi Daniel,

A couple thoughts:

- Are you worried about different velocities between subjects or within subjects? If the design is a repeated measures (pre vs. post on each subject), then you may not have to worry much about variance in speeds between subjects. If I remember right, it can actually be bad for repeated measures if the between-subjects variance is very small.

- I would hesitate to scale the free moment (peak?) by velocity unless you have evidence that it increases linearly with increasing velocity. If the change in velocity is a natural effect of the intervention (vs. an error in the experiment) then you probably would not want to scale out its effect. If there's big differences in velocities pre and post (e.g. some of the subjects sped up, some slowed down), I would first ask why they responded that way before trying to account for it statistically. Hard to say without know what your intervention was.

Good luck,
Ross

[Petcu Daniel]

Re: velocity normalization

Dear Mr. Ross,

Thank you for your answer !
We are trying to evaluate the effect of a knee surgery intervention on the free moment using one AMTI AccuGait force platform. The design is a repeated measures one - pre vs. post on each subject.
One big problem is that we don't have [yet] a valid system for measuring the velocity so we are comparing the pre and post operative peak values looking at stance time duration [as this is decreasing non-linearly with walking speed, http://www.sciencedirect.com/science...66636296806552 ]. The difference in stance time duration between pre and post operative trials is, for some subjects, between 15-20% [the speed postoperatively is smaller than preoperatively, which is normal] but the peak free moment is opposite : smaller preoperative than postoperative. We have made a preliminary verification, with one health subject, for the relation between peak free moment and stance time duration using a metronome. The results was a increase of peak free moment with decreasing of the stance time duration - similar with the case of subjects with a surgical intervention. So, from our preliminary data it seems that free moment increase with increase of stance time duration [respectively with decrease with velocity]
Now I'm confused because :
1- why we have obtained an inverse relation between peak of free moment and velocity ?,
2- as the change in velocity is a natural effect postoperatively how can I distinguish between intervention and velocity's effects with such type of design experiment ?

Thank in advance for your help,
Respectfully,
Daniel

[Gannon White]

Re: velocity normalization

As you were replying I was thinking what you expressed in #2. Anxious to hear what you come up with.

Regards,
Gannon

[Sjoerd Bruijn]

Re: velocity normalization

Hi Daniel,
another way would be to not normalize your variable, but statistically model the effect of walking speed. With an ANCOVA, you could use walking speed as covariate, however, this would not allow you to adress the change in walking speed (unless you use the change in walking speed as a covariate). The way I have done this in previous studies, is to use Generalized Estimation Equations, which is basically a regression technique with correction for repeated measures. Then, you can include walking speed as a covariate (for an example (without intervention though), see for instance http://www.sciencedirect.com/science...91422213000693 ). If you have a control group that walks at a range of speeds that is within the domain of the patients pre- and post-, you would be able to see (amongst other things)
- if there is an effect of walking speed in general
- if there is a different effect of walking speed in patients vs controls (interaction effect between group and speed)
- if there is a different effect of walking speed in patients pre- vs post- (interaction effect between intervention and speed)
- direct effects of group, intervention, etc, etc.
Hope this helps, best,
Sjoerd

[Petcu Daniel]

Re: velocity normalization

The results was a increase of peak free moment with decreasing of the stance time duration - similar with the case of subjects with a surgical intervention.

I've to make an erratum : "The results was an increase of the free peak moment with the increase of the stance time duration...".
Which means that peak free moment has increased with the decrease of walking velocity. This is opposite with tendency of increasing of peak free moment with the increase of velocity [ http://www.sciencedirect.com/science...50641191900376 ]
I apologies for this error !
Thank you for your understanding!
Respectfully,
Daniel

[Petcu Daniel]

Re: velocity normalization

Thank you for your answers !
I have a lot to study !
Respectfully,
Daniel

[Ton van den Bogert]

Re: velocity normalization

Sjoerd Bruijn's posting reminded me of a recent review article that had some critical comments on the use of ANCOVA to deal with effects of walking speed.
See: http://www.ncbi.nlm.nih.gov/pubmed/22019141

Ton van den Bogert

[Petcu Daniel]

Re: velocity normalization

Sjoerd Bruijn's posting reminded me of a recent review article that had some critical comments on the use of ANCOVA to deal with effects of walking speed.
See: http://www.ncbi.nlm.nih.gov/pubmed/22019141

Ton van den Bogert

Dear Members,
This is why I like the Einstein's quote : "No amount of experimentation can ever prove me right; a single experiment can prove me wrong" !
Nice weekend,
Respectfully,
Daniel

[Petcu Daniel]

Re: velocity normalization

- I would hesitate to scale the free moment (peak?) by velocity unless you have evidence that it increases linearly with increasing velocity.
Good luck,
Ross

Dear Members,
I would like to ask you for some references regarding the influence of walking velocity on the peak free moment ! I would like to compare these data with the preliminary data resulted from our measurements.
Thank you in advance !
Respectfully,
Daniel

[Petcu Daniel]

Re: velocity normalization

Dear Mr. Almosnino

Thank you for the references !
In Li et al (2001) the relationship between free moment and speed is treated only qualitatively, even from the graphs seems to be some significant quantitative differences between different walking conditions ! In "The free moment in walking and its change with foot rotation angle" the walking velocity doesn't significantly differ between tested conditions. But the conclusion is very interesting as the externally foot rotation angle seems to have a significant influence on peak normalized free moment, compared with normal walking. I haven't seen the foot rotation angle taken into in account in other articles treating the free moment ! The very interesting reference indicated by Mr. Ton van den Bogert raise some important problems, as "even statistical difference in a gait variable between groups can either partially or completely be due to differences in other related variables, such as speed" !

Sincerely,
Daniel