Multivariate RM ANOVA for amplitude and temporal effects in EMG

[k-bill]

Hi Todd,

I’m looking for the right statistical procedure to answer my research question.

I have EMG data (4 different muscles) of ~45 participants. Each participant performed three different types of sidestep cuts. I would like to analyze the time spans from 100 ms before touchdown until toe-off.

I read your paper on simultaneously assessing amplitude and timing effects and was wondering

• if there is a way to conduct a multivariate repeated-measures ANOVA to simultaneously assess these effects in my EMG signals, or is there an alternative you could suggest?

If it is possible to conduct the multivariate repeated-measures ANOVA to assess temporal and amplitude effects simultaneously,

• would it be problematic for the registration that the time before touchdown until ground contact is the same for all participants (100 ms) but from touchdown to toe-off it is not? Should I rather choose fixed values, e.g, 100 ms before touchdown until 100 ms after touchdown?

As an additional question, as I’m not interested in between-subject effects,

• would you perform the statistics on the raw signal or on the processed signal?

Thank you very much for everything and all the best,

Kevin

[0todd0000]

Hello!

• if there is a way to conduct a multivariate repeated-measures ANOVA to simultaneously assess these effects in my EMG signals

SPM can indeed be used to do this, but this is unfortunately not yet supported in spm1d.

or is there an alternative you could suggest?

As an alternative you can skip to post hoc tests and use paired Hotelling's tests like in this example

If it is possible to conduct the multivariate repeated-measures ANOVA to assess temporal and amplitude effects simultaneously,

Theoretically, yes, but I am not aware of any software package that supports this. The closest I know is FDASRSF but I don't think it supports multivariate data.

Your last two questions (regarding registration and raw vs. processed data) are tough to address here because I believe that they are empirical questions. In general I think it is reasonable to try analyses in as many ways as possible (and feasible) and to compare the results. Since this type of analysis is not directly supported in spm1d these questions are a bit difficult to address in this forum.

[k-bill]

Thank you very much, I appreciate it a lot.