We have measured EMG during walking in two different population, and would like to compare their muscle activity.
While some former studies have used statistical inference (i.e. t-tests for max values), we would like to assess the whole time series. Our aim is to show difference in or near specific events of the gait cycle.
Any advice regarding how to perform a statistical inference for time series? or where should I obtain more information regarding this topic?
Thank you!
I would recommend you to have one or some hypothesises of why or what do you think about the difference between these two population, and to make assumptions in detail. this could be from you observation, some one's working experience or unconfirmed theory ......
Then think about what kinds of EMG data being used to confirm your assumption..... e.g. if you think the difference of the two population is due to kinds of fatigue, then you can choose median frequency
Definitly no data mining, as with a time series EMG, you can generate tons of variables.....
I think that the T1 scale is a good statistical tool for assessing parameter changes over time.
- DOI: 10.1016/j.poamed.2016.03.001
You should also consider in what domain you intend to establish the comparison (i.e. raw time series or extracted features from raw time series). If you're interested in performing feature extraction from EMG I would recommend using Time Series Feature Extraction Library (TSFEL) (https://github.com/fraunhoferportugal/tsfel), as it includes a wide range of feature extraction methods implemented.
For the normalized envelope of the EMG signal, I would recommend SPM 1d technique (http://www.spm1d.org/).
In addition, if you're interested in percentage difference (like symmetry) you can try Symmetry Function tool
( Article A New Method of Evaluating the Symmetry of Movement Used to ...
) or (Article A New Method of Evaluating the Symmetry of Movement Used to ...
).Cheers!
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series