[Eeglablist] Random selection of trials

David Groppe dgroppe at cogsci.ucsd.edu
Sat Apr 3 10:36:16 PDT 2010


On Fri, Apr 2, 2010 at 4:09 PM, Tim Curran <Tim.Curran at colorado.edu> wrote:

>
> On Apr 2, 2010, at 3:39 AM, Kris Baetens wrote:
> > @ Tim curran: I have read Luck's book and I clearly see why using
> different trials in case of
> > peak measures (which we do not use) might be problematic. However, I
> can't quite understand why
> > average amplitudes would be 'immune' to this sort of problem; I think
> sensitivity to this
> > problem decreases as the number of time points in your average window
> increases? (In extremis:
> > an average amplitude of two time points would, in the same logic as he
> discribes, still be extremely
> > vulnerable, no?)
>
> I had the same question/doubt when I first read this.  I had long email
> discussions with Luck about it, and actually ended up writing a simulation
> in order to convince myself that he was right.  The bottom line is that I
> cannot explain it myself, but I became convinced that he is correct.
> Tim
>
>
Here's the rationale:
1) When you have fewer trials in one condition than another, the variance of
the estimated ERP is greater.  Thus you're more likely to get extreme values
(both above and below the true mean) in the condition with fewer trials.

2) If you measure the mean amplitude in a particular time window, the
extreme above-the-mean and below-the-mean values tend to cancel out.  So, on
average, your estimates of mean amplitude across the two conditions won't
differ (if they really have the same mean amplitude).

3) If you measure the peak amplitude in a particular time window, you're
measuring the most extreme value of a particular polarity.  Now the extreme
values in one direction (e.g., above the mean) will NOT be canceled out by
extreme values in the other direction (e.g., below the mean) and you'll be
more likely to get bigger peak amplitudes in the condition with fewer trials
(even if they really have the same peak amplitude).

    hope that helps,
          -David




>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20100403/09320704/attachment.html 


More information about the eeglablist mailing list