Page 1 of 1

Offline analysis two electrodes

Posted: 19 Aug 2014, 12:03
by Ramiro
Hello, while performing topography with two electrodes and five runs I get the following error:
Not enough unique points specified sample

Anyone know what could be the problem?

Ramiro HernĂ¡n Gatti

Re: Offline analysis two electrodes

Posted: 20 Aug 2014, 11:06
by boulay
I'm sorry I don't understand.

What data are you using? How was it collected? What was the task?
What application is generating that error?
What are you trying to get the application to do? (Note, I don't think 'topography' is correct in this instance)

Re: Offline analysis two electrodes

Posted: 21 Aug 2014, 10:12
by Ramiro
Hello, let me explain a little better.
We work with the motor imagery paradigm in order to record brain activity. For this we use two electrodes through g.mobilab plus device recorded the activity against different tasks, such as opening and closing the hands. The idea is to observe the Coefficient of Determination in Topography Plots.
We have 200 samples of each task, and when we do the Topography Plots in the BCI2000 Offline Analysis tool we get the error:
Not enough unique points specified sample

Clarify, we modify the file eloc so that only two positions are both we use.

Thank you.

Re: Offline analysis two electrodes

Posted: 21 Aug 2014, 15:20
by boulay
I don't use the offline analysis tool, nor do I understand its underlying algorithms, but 2 points are not enough to build a 2-dimensional surface. That may be the reason for the error.

Are you unable to get the coefficients without generating a topography?

Re: Offline analysis two electrodes

Posted: 21 Aug 2014, 22:12
by Ramiro
Got absolutely right, I have only two points.
Yes, I can see in the Spectra Plot. Only had the doubt. Thank you.

On the other hand, what tool do you use to analyze the signal?

Re: Offline analysis two electrodes

Posted: 21 Aug 2014, 23:43
by boulay
I use many different things. I can't really recommend one over the other. It really depends on the situation.

BCI2000 -> Matlab then custom code and other toolboxes (e.g., chronux). I usually do this for small-ish tasks.
BCI2000 -> EEGLAB (a Matlab toolbox). EEGLAB has a plugin to load BCI2000 dat files. I like EEGLAB's way of managing and doing stats on large datasets... but it is difficult to extend to custom analysis pipelines. But it comes with quite a few very common and very useful analysis pipelines.
BCI2000 -> EEGLAB -> brainstorm . Brainstorm is great for doing source imaging. I find it very difficult to write custom code for because its codebase has several layers of java and gui abstractions before you can find the actual data manipulation functions.
BCI2000 -> Python using BCPy2000. Despite the website not having an entry since mid 2011, the code itself is actively maintained. I use Python+BCPy2000 when I'm experimenting around with different machine learning techniques.