Is it a good idea to never do "Manuel trial rejection" because of the possibility of the researcher biases coming into play? For example, some question Ancel Keys nutrition work on saturated fat because he took out countries that told a different story that most of the others (something along those lines).
Good question, and it's been debated in the field for about 30 years now. The problem is that no algorithm or combination of algorithms does a satisfactory job. It is important to do data cleaning in a blind fashion. But fortunately, EEG data are sufficiently noisy and variable that it's impossible to predict whether rejecting a given trial will have a systematic bias on the direction of the final outcome of the study. That's very different from looking at the final results and then going back to remove data, which is obviously unethical.
Resting state data is nearly the same. You just don't have a time=0 event to time-lock to. It's still a good idea to epoch the data into, e.g., 2-second segments. That can facilitate cleaning and is used for spectral analysis, e.g., using Welch's method. I don't have any plans on making a video about microstates, but I'm sure there are other videos on that somewhere...
You take the average signal across electrodes (or all electrodes that aren't bad, as he recommends), and substract that from all channels to remove common noise in all channels It's typicall called "Common Average Referencing", look that up for more information
I just wish preprocessing hadn’t been labelled as boring for newcomers to the field - but perhaps people coming into EEG from signal processing is a minority who find the preprocessing choices and how it can impact and bias analysis exciting 😅
True, I actually did enjoy preprocessing in the beginning. It became tedious after many years. But you're right that choices made during preprocessing impact the subsequent analyses, so it's a super-important aspect of working with the data regardless of your emotional experience with it :P
Is it a good idea to never do "Manuel trial rejection" because of the possibility of the researcher biases coming into play? For example, some question Ancel Keys nutrition work on saturated fat because he took out countries that told a different story that most of the others (something along those lines).
Good question, and it's been debated in the field for about 30 years now. The problem is that no algorithm or combination of algorithms does a satisfactory job. It is important to do data cleaning in a blind fashion. But fortunately, EEG data are sufficiently noisy and variable that it's impossible to predict whether rejecting a given trial will have a systematic bias on the direction of the final outcome of the study. That's very different from looking at the final results and then going back to remove data, which is obviously unethical.
@5:35 Analyzing data with HTML is suicide. HTML is not even a programming language.
Don't worry, Phil, it was just a stupid joke ;)
Hi, Mike. How preprocessing steps differ to resting state data? And are you planing to make a video about EEG microstates?
Resting state data is nearly the same. You just don't have a time=0 event to time-lock to. It's still a good idea to epoch the data into, e.g., 2-second segments. That can facilitate cleaning and is used for spectral analysis, e.g., using Welch's method.
I don't have any plans on making a video about microstates, but I'm sure there are other videos on that somewhere...
5:27 I paused the video and honestly was trying to imagine HTML environment for analyzing the neural data....haha
Just a little joke to make sure you're paying attention :P
I don't clearly understand average references, what do you are meaning by that?
You take the average signal across electrodes (or all electrodes that aren't bad, as he recommends), and substract that from all channels to remove common noise in all channels
It's typicall called "Common Average Referencing", look that up for more information
I just wish preprocessing hadn’t been labelled as boring for newcomers to the field - but perhaps people coming into EEG from signal processing is a minority who find the preprocessing choices and how it can impact and bias analysis exciting 😅
True, I actually did enjoy preprocessing in the beginning. It became tedious after many years. But you're right that choices made during preprocessing impact the subsequent analyses, so it's a super-important aspect of working with the data regardless of your emotional experience with it :P
awesome. thank you
No, you're awesome!
😂 u really dont like preprocessing data