[Eeglablist] About the analysis result

[Scott Makeig]

Note that averaging across ERCOH (crossf) outputs assumes that they are 
equivalent. Often, researchers average similar channel locations across 
subjects, though they do not know whether the same classes of EEG 
sources project in the same proportions to a given channel location 
(e.g., Cz) or not.

An attractive possibility is to average over similar independent 
components from all subjects, or from all subjects who exhibit 
components of a certain category. The problem then is how to categorize 
components so as to identify them across subjects. We have approached 
this problem from different directions - there is no one best answer, I 
believe.

Currently, EEGLAB users are 'on their own' on this - we have no 
component clustering functions included, though this is a future goal.

There might eventually be several component clustering plug-ins to 
EEGLAB, as there are many approaches to clustering. This might be a good 
area for open source contributions.

Scott Makeig

Arnaud Delorme wrote:
>>
>> *1. what is the meaning for the blue and green curve in the figure? Is 
>> it the similar meaning with channel time-frequency function? their 
>> axis label seems different.*
> 
> The meaning of the curves is describe in the crossf and timef function 
> help (this is true only in EEGLAB 4.2). It is different for timef and 
> crossf. From crossf (from the help message):
> 
>    Assuming both 'plotamp' and 'plotphase' options are 'on' (=default), 
> the upper panel
>     presents the magnitude of either phase coherence or linear 
> coherence, depending on
>     the 'type' parameter (above). The lower panel presents the coherence 
> phase difference
>     (in degrees). Click on any plot to pop up a new window (using 
> 'axcopy()').
>     -- The upper left marginal panel shows mean coherence during the 
> baseline period
>        (blue), and when significance is set, the significance threshold 
> (dotted black-green).
>     -- The horizontal panel under the coherence magnitude image 
> indicates the maximum
>        (green) and minimum (blue) coherence values across all 
> frequencies. When significance
>        is set (using option 'trials' for 'boottype'), an additional 
> curve indicates the
>        significance threshold (dotted black-green).
> 
>> ** *2. What is the meaning of BootStrap Signification value in it? How 
>> will it affect the results?*
> 
> Computing a bootstrap distribution means that the (timef or crossf) 
> function will apply the same formula it used to compute ERPS, ITC or 
> coherence to the shuffled data (shuffled across time, across trials or 
> both). For instance in timef, the data might be shuffled 200 times and 
> the ERSP (Event-related Spectral Power) will be computed for each 
> shuffling. You will end up with a bell-like distribution of ERSP values 
> and if the non-shuffled ERSP "belongs" to this distribution (i.e. it is 
> not in the 5% lowest or highest tails for instance), the ERSP (at this 
> time point and frequency) is considered significant at 5%. These 
> non-parametric statistics do not assume a probability distribution model 
> for the measure (unlike a t-test for instance).
> 
>> *3. How can I save the results of channel cross-coherence function in 
>> a file for later use?*
>> **
> 
> If you are using the graphical interace. After calling the pop_timef for 
> instance (menu "Plot > Time-frequency transforms > Channel 
> time-frequency"). Type "h" on the command line to show the history
> 
> "figure; pop_timef( EEG, 1, 1, [-1000  2000], [3         0.5] ,'type', 
> 'phasecoher', 'topovec', 1, 'elocs', EEG.chanlocs, 'title','Channel FPz 
> ERSP and ITC','padratio', 4, 'plotphase', 'off');"
> 
> copy and paste and add some outputs to the function:
> 
> figure; [ersp,itc,powbase,times,freqs,erspboot,itcboot] = pop_timef( 
> EEG, 1, 1, [-1000  2000], [3         0.5] ,'type', 'phasecoher', 
> 'topovec', 1, 'elocs', EEG.chanlocs, 'title','Channel FPz ERSP and 
> ITC','padratio', 4, 'plotphase', 'off');
> (for details about the output see timef function help).
> 
> then save the outputs:
> 
> save -mat myfile.mat ersp itc powbase times freqs erspboot itcboot
> 
> these outputs may be combined for several subjects and used in the 
> command line function tftopo() (no menu equivalent or graphical interface).
> 
>>  *4. (some out topic) Suppose, I have several subject EEG data stored 
>> in different files. What post-process I need to do after finished the 
>> channel cross-coherence function one by one for all subject?*
> 
> You may use tftopo above to combine all the coherences (note that tftopo 
> is made to process timef outputs not crossf outputs, so the units on the 
> figure might have to be slightly modified).
> 
> Best
> 
> Arno
> 
> 
> -- 
> 
> *Arnaud Delorme, Ph.D.*
> Computational Neurobiology Lab, Salk Institute
> 10010 North Torrey Pines Road
> La Jolla, CA 92037 USA
> 
> *Tel* : /(+1)-858-458-1927 ext 15/
> *Fax* : /(+1)-858-458-1847/
> *Web page *: www.sccn.ucsd.edu/~arno <http://www.sccn.ucsd.edu/%7Earno>
>