[Eeglablist] Statistically determining earliest time point of difference between two single trials

[Eric Fields]

Hi Kaelasha,

What do you mean by " looking at single trials"? There is really no way to
determine whether or when two individual trials differ, both because signal
to noise is very low in individual trials and because inferential
statistics require multiple trials to get a measure of variability.

If you mean you want to determine when two conditions begin to differ
averaged across trials, this is actually a difficult problem to solve. The
tmax/Fmax approach that Steve suggests will significantly underestimate how
early differences actually begin. See:

https://onlinelibrary.wiley.com/doi/10.1111/psyp.13468

Simulations suggest that cluster-based procedures may actually give a more
accurate estimate of effect onset, but they do not offer error rate control
for individual time points and don't accurately reflect timecourse in all
situations.

Those approaches consist of calculating a test at every time and seeing
where effects begin. Approaches to obtaining an estimate of onset latency
directly are reviewed by Kiesel et al.:

Kiesel, A., Miller, J., Jolicœur, P., & Brisson, B. (2008). Measurement of
ERP latency differences: A comparison of single-participant and
jackknife-based scoring methods. Psychophysiology, 45(2), 250-274.
https://doi.org/10.1111/j.1469-8986.2007.00618.x

Eric

-----
Eric Fields, Ph.D.
Postdoctoral Fellow
Cognitive and Affective Neuroscience Laboratory
<https://www2.bc.edu/elizabeth-kensinger/>, Boston College
Aging, Culture, and Cognition Laboratory <http://www.brandeis.edu/gutchess/>,
Brandeis University
eric.fields at bc.edu


On Tue, Sep 3, 2019 at 3:50 AM Stephen Politzer-Ahles <
politzerahless at gmail.com> wrote:

> Hi Kaelasha,
>
> This is described in the paper by Groppe and colleagues (2011):
> https://onlinelibrary.wiley.com/doi/full/10.1111/j.1469-8986.2011.01273.x.
> Crucially, if you want to know when an effect starts to be significant,
> you'll want to use the test with strong control of family-wise error rate
> (the non-cluster-based test), not the cluster-based test (which only has
> weak control of family-wise error rate, and doesn't license inferences
> about when an effect starts to be significant).
>
> This test is implemented in the Mass Univariate ERP Toolbox, which works
> with EEGLAB.
>
> Best,
> Steve
>
> ---
> Stephen Politzer-Ahles
> The Hong Kong Polytechnic University
> Department of Chinese and Bilingual Studies
> http://www.mypolyuweb.hk/~sjpolit/
> <http://www.nyu.edu/projects/politzer-ahles/>
>
>
> On Mon, Sep 2, 2019 at 4:25 PM Kaelasha Tyler <kaelasha.tyler at gmail.com>
> wrote:
>
> > Hi all,
> >
> > I am wanting to determine the earliest time point at which a
> statistically
> > significant deviation between two matched trial ERPs can be detected,
> > looking at single trials.
> >
> > I have looked across this discussion forum but haven't seen any thread
> > that actually explains how to do this (if possible) in EEGlab.
> >
> > Any pointers anyone?
> >
> > I am replicating a study which used sample-wise truth-label switching
> Monte
> > Carlo permutation tests. However for my purposes it wouldn't have to be
> the
> > exact same statistical procedure as in this study, as long as I can
> achieve
> > the same statistical ends.
> >
> > Thanks in advance,
> >
> > Kaelasha
> > _______________________________________________
> > 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
> >
> _______________________________________________
> 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
>