Statistical Parametric Mapping (SPM) Analysis

[Eoin Doyle]

Hi All.

We are doing statistical parametric mapping (SPM) analysis for biomechanical analysis research. Our research group has previously used SPM to compare acceleration signals from inertial measurement units (IMU) (https://doi.org/10.1080/02640414.2019.1692997).

Has anyone used SPM to compare signals using different measurement scales (eg. m/s/s and BW/s). Assuming the data will need to go through data registration (sync of peaks) and potentially normalised from min to max.

Thanks in advance.

Regards,
Eoin Doyle

[Samuel Acuna]

Interesting idea. What exactly is the question you are trying to answer? SPM is just a method for doing statistics across a time series (e.g., doing a t-test at each percentage of the gait cycle while accounting for multiple comparisons). So, if you wouldn't do a t-test between data using different measurement scales, I don't see justification for doing SPM using different measurement scales. However, if you are doing some sort of regression between measurements using SPM, there is a little bit of literature out there on it, but it is not as popular because it is difficult to interpret.

If you are just trying to compare the shape of curves, there are some alternative approaches, like the cosine similarity index. If you are trying to quantify the curves, you might find success using Functional Data Analysis (https://www.psych.mcgill.ca/misc/fda/index.html).

If you just want to show that the measures are very similar, perhaps just finding the correlation between the group-average signals is sufficient.