# Time Warping/Aligning Events from different epoch lengths

Dear community,
for my thesis i conducted a motor sequence learning task during which participants had to "dance" a certain sequence on a dancing mat. One sequence consists of six responses (steps). All of their steps were recorded. The participants were rather slow in the beginning, taking about 5 seconds to perform the whole sequence. After practicing for some time, they could perform the whole sequence within 500ms.
For my analysis i would like to epoch the time window during which they responded. Since they became faster over time, this will result in different epoch lengths. One idea i had was to use dynamic time warping to be able to align the steps. Is there any way to do this in brainstorm? Does dynamic time warping make sense when the span of response time reaches from 500ms to 5s? Or are there any alternatives or suggestions how to go about the different epoch lengths?
Here i have some screenshots of my data. As you can see i have markers for each step, but they are not aligned and i would like to align them. The hypothesis i have is that there might be a increase in theta power around the 3rd or 4th step (concatenation point, loading the next steps). But if i just average the response time window using an epoch with the longest response time, i run into the problem that after a while most steps occur in the first second.
I have markers which indicate the start and end of each response, however you cannot see them in the screenshot here.

I would appreciate any ideas.
Thank you for your time, Victoria

Thanks for reaching out. Unfortunately, there is no time-warping tool in Brainstorm. If your hypothesis involves an increase in theta power around specific steps of the sequence, you could epoch your data around each step and compute the power changes around them. However, a potential issue is that as the motor sequence shortens with practice, the inter-step interval may become shorter than one cycle of theta oscillation (125-250 ms), making event-related modulation of theta potentially meaningless.

An alternative approach could be to derive a regressor time series from the actual step sequences performed in each trial, and then use this against the brain time series to identify which regions show greater consistency with the behavior over time (for instance).