How to analyse movement variability???

[Patrick Fürst]

Hi everyone. I'd like to evaluate the intrapersonal movement variability in repeated cutting movements during and shortly before ground contact and am currently thinking about ways to calculate movement variability of kinematic, kinetic and electromyographic data. The aim is to compare different subjects in their movement variability.

One thing is to look at the variability of a certain parameter at a certain time point, e.g. knee flexion at initial contact. To do this, I think we could use the coefficient of variation (CV).

Another thing is to look at the variability of whole curves, i.e. to look at how the knee flexion curves over the whole ground contact phase of the cutting leg differ between trials. Here I thought of using the 'adjusted coefficient of multiple determination', which has been described by Kadaba et al. (1989) as a means to measure the 'similarity of waveforms'. I think it is a good tool to analyse variability, but in the end you do not know where the variability comes from (i.e. from the beginning of the curve or the end of the curve, from a big difference over a short time period or from a smalll difference over the whole time period...).

Do you have any comments to my ideas and thoughts or suggestions which other ways there are to analyse movement variability?

Thanks!
Patrick

[Robert D. Catena]

Re: How to analyse movement variability???

I am not familiar with Kadaba's method, but it sounds similar to the Pearson Product-Moment correlations described in Stergio's Innovative analyses book (Great book by the way, but I hope there is new addition soon).

One thought to your dilemma in finding the time-frame variability is to run a moving envelope correlation analysis. The specifications (size) of this envelope should be based off of pilot data and not your real data as to not bias it.

[Mark Halaki]

Re: How to analyse movement variability???

Hi Patrick,
We have recently published a method to separate offset and waveform variability in curves (kinematic or EMG) which might be useful:
O'Dwyer, N., Smith, R., Halaki, M., & Rattanaprasert, U. (2009). Independent assessment of pattern and offset variability of time series waveforms. Gait & posture, 29(2), 285-289. doi: 10.1016/j.gaitpost.2008.09.005
Regards,

Mark Halaki, PhD
Senior Lecturer
Discipline of Exercise and Sport Science
Cumberland Campus C42
The University of Sydney
PO Box 170
Lidcombe NSW 1825
AUSTRALIA
Tel: +61 2 9351 9883
Fax: +61 2 9351 9204

[Floris van Vugt]

Re: How to analyse movement variability???

Hi everyone,
I developed a (very simple) model to analyse within-subject variability in a discrete timing task, segmenting variability into trial-to-trial variability on the one hand, and "consistent" temporal deviations on the other. We used the method to study chunking in serial production tasks. For this reason, I'm not sure whether it is relevant in this discussion, but just in case, here is the paper:

van Vugt FT, Jabusch HC and Altenmueller E (2012). Fingers phrase music differently: trial-to-trial variability in piano scale playing and auditory perception reveal motor chunking. Front. Psychology 3:495. http://www.frontiersin.org/Auditory_...00495/abstract

Regards,
Floris

[Mark Corbett]

Re: How to analyse movement variability???

You should only use CoV for heteroscedastic variables (where variability increases as measured values increase). This is typically not that case for anatomical angles. Hopkin's typical error, or Bland & Altman's limits of agreement might be more appropriate.

Kadabas coefficient of multiple correlation is fairly straightforward and works across an entire kinematic cycle. However like most correlation coefficients it just gives a number up to 1. It's difficult to judge whether a 0.75 is adequate, relevant or much better than a 0.6
Richard Baker advocates that to be meaningful, interpretable and useful movement variability needs to be reported back in the original measurement units. Most clinicians would agree that <1 degree of variability would be excellent, ~5 degrees marginal. To do this with his Gait reliability profile he simply takes the mean angle across the cycle and analyses the test-retest variance both inter-trial and inter-session.

[Ezio Preatoni]

Re: How to analyse movement variability???

Dear Patrick,

the answer to your question is not straightforward because the method you may choose depends on many factors, among which the aims of your study (e.g. are you trying to identify: pathological vs. healthy or small changes over time due to an intervention? changes in a specific feature or changes in coordination between different elements?) and the characteristics of the measures you are collecting. There are quite a few good papers on different techniques to analyse MV, both with discrete and continues measures. Besides what has already been suggested, a non-comprehensive list to have a start with the different approaches may be:
- Chau, T., Young, S., & Redekop, S. (2005). Managing variability in the summary and comparison of gait data. Journal of NeuroEngineering and Rehabilitation, 2(1), 22.
- Duhamel, A., Bourriez, J. L., Devos, P., Krystkowiak, P., Destée, A., Derambure, P., et al. (2004). Statistical tools for clinical gait analysis. Gait & Posture, 20(2), 204-212.
- Hamill, J., van Emmerik, R. E. A., Heiderscheit, B. C., & Li, L. (1999). A dynamical systems approach to lower extremity running injuries. Clinical Biomechanics, 14(5), 297-308.
- Ryan, W., Harrison, A., & Hayes, K. (2006). Functional data analysis of knee joint kinematics in the vertical jump. Sports Biomechanics, 5(1), 121-138.

Finally, I and some colleagues have recently published a review paper that tries to address the issue of MV from many different perspective, with a particular focus on sports movements:
- Preatoni, E., Hamill, J., Harrison, A. J., Hayes, K., Van Emmerik, R. E. A., Wilson, C., et al. (2012). Movement variability and skills monitoring in sports. Sports Biomechanics [Epub ahead of print] - http://dx.doi.org/10.1080/14763141.2012.738700.

Best wishes
-Ezio

--
Dr Ezio Preatoni
E.Preatoni@bath.ac.uk

Lecturer in Biomechanics and Motor Control
Sport, Health & Exercise Science
Department for Health | University of Bath

[Julius Verrel]

Re: How to analyse movement variability???

Spatiotemporal morphable models (Giese & Poggio, 2000) may be interesting. The allow to separate temporal and spatial aspects of variability. Publications can be found on Martin Giese's lab's website:

http://www.compsens.uni-tuebingen.de/joomla/index.php?option=com_jresearch&view=publicationslist&Itemid=65&lang=en&limitstart=125

Best, Julius

[Suzi Edwards]

Re: How to analyse movement variability???

Another method that you may want to look at is Hopkins, W. G., & Hewson, D. J. (2001). Variability of competitive performance of distance runners. Medicine & Science in Sports & Exercise, 33(9), 1588-1592.

[Brian S. Sinner]

Re: How to analyse movement variability???

Hi everyone. I'd like to evaluate the intrapersonal movement variability in repeated cutting movements during and shortly before ground contact and am currently thinking about ways to calculate movement variability of kinematic, kinetic and electromyographic data. The aim is to compare different subjects in their movement variability.

One thing is to look at the variability of a certain parameter at a certain time point, e.g. knee flexion at initial contact. To do this, I think we could use the coefficient of variation (CV).

Another thing is to look at the variability of whole curves, i.e. to look at how the knee flexion curves over the whole ground contact phase of the cutting leg differ between trials. Here I thought of using the 'adjusted coefficient of multiple determination', which has been described by Kadaba et al. (1989) as a means to measure the 'similarity of waveforms'. I think it is a good tool to analyse variability, but in the end you do not know where the variability comes from (i.e. from the beginning of the curve or the end of the curve, from a big difference over a short time period or from a smalll difference over the whole time period...).

Do you have any comments to my ideas and thoughts or suggestions which other ways there are to analyse movement variability?

Thanks!
Patrick

Patrick,

Could you do something like Cormie et al (2008). They compared the force-time, power-time, and velocity-time curves during squat jumps with different loads. They first normalized the curves by re-sampling each subjects trials to ouptut a constant number of samples (i.e. 500), then statistically analyzed them which gave them the areas of the curve where there were differences [e.g. shaded regions in Figure 3]. I don't know if this could somehow be applied to assessment of of variability.

Cormie. P, McBride. J, & McCaulley. G. (2008). Power-time, force-time, and velocity-time curve analysis during the squat jump:Impact of load. Journal of Applied Biomechanics, 24, 112-120.



Brian

[Patrick Fürst]

Re: How to analyse movement variability???

Hello everyone,
I would like to thank you all for your valuable comments. I found many of the suggested readings very helpful and your advice sure helped me in the preparation of my study.
Thank you and best wishes,
Patrick