Hi Mike and everyone,
I hope you are doing well, and I am so sorry if this is a very silly question!
I am designing a study for a project where we want to look at toddlers’ frontal power asymmetry in alpha. I realise this is most often studied from the baseline data, where a child sits on their parent’s lap and watch something quietly (e.g., a video clip, experimenter showing a toy) for 1-5 minutes.
But my PI wants to use a presentation of static images to get the asymmetry score. I could present the stimlus (image) for 1-2 seconds, but I don’t think I could make it longer than a few seconds as the baby won’t keep looking at it! So the presentation would be more like an ERP study, even though we will looking at power.
I was wondering how I should analyse the data. I was thinking about 2 options:
- Use a time-frequency analysis, and get the asymmetry focusing on the power in a specific time window (although there’s no other study which has done this, so no strong a priori hypothesis as to which time window to look at!).
- Do a frequency analysis using a shorter time window, e.g., 200ms, rather than conventional 1s!?
- I came across this event-related spectral pertubation (ERSP), which looks promising and has been used for infant data at least once - but I wonder how commonly it is used? I don’t think I remember it being mentioned in your time-series data analysis book and tutorials (although I must admit it was a few years ago I read the whole book)?
Again so sorry if it’s a silly question, but it’d be great if I could have some advice!
Hi Saya. Sure, you can compute frontal alpha asymmetry in 1-2 seconds. I think the question is whether that’s enough time for a reliable and robust estimate. I don’t know the answer, but I guess if you have a lot of trials, it should be OK.
ERSP is the same thing as time-frequency power, so yes, I mention that in the book
You might also consider applying the “standard” frontal alpha asymmetry analysis approach. It’s usually computed using resting-state data, but the method can of course be applied to task-related data. Do you have a separate resting-state-like recording in the same infants that you could use for comparison?
Thank you very much for your response, and so sorry I must have missed the ERSP in your book!! I’ll go back and check again.
It’s with 18-month-old children so hopefully we’ll get at least 10 good trials per condition. In infancy research, 10 trials per condition is considered to be okay if not the best (although for adult research that might look too few!). If I do a frequency analysis using Hanning windowing for example with 50% overlap, would it be like using a 125ms-window?
Yes we’ll be collecting resting state data too, which will hopefully be at least 1 minute long.
I thought a “standard” approach would be using a Fourier transform with some kind of windowing (e.g., Hanning)?
Yeah, that sounds right. You can look up some resting-state papers to see their methods. It would also be worth looking around on pubmed/google-scholar to see whether other people have computed brief task-related alpha asymmetry. I’m not familiar with any studies like that off-hand, but that isn’t my field, so it’s possible that such studies exist but I don’t know of them.
Yes I’ve seen one paper that did a task-related alpha asymmetry with infants, but it might be worth looking at adult papers too. Thank you very much Mike, it’s very kind of you to respond to my questions !