[FieldTrip] cluster-level statistic effect size

Eric Maris e.maris at psych.ru.nl
Sat Apr 27 13:04:21 CEST 2013


Dear Mark,



When computing cluster-level statistics recent versions of Fieldtrip now
also compute the SD and confidence interval of the estimated p-value for
each cluster. But I was wondering if it is also possible to compute an
effect size such as Cohen’s d of the observed effect. Normally, 2*t-value
divided by sqrt(df) gives Cohen’s d. However, this does not work for the
summed t-values in each cluster.

Any suggestion how to compute an effect size for the cluster-level
statistic?



Why would you want to compute an effect size for a neuronal variable (such
as an evoked response, an induced oscillation)? An effect size for a
behavioral variable (such response time or accuracy), on the other hand,
would definitely make sense. In cognitive neuroscience, we use neuronal
variables to explain behavior; it’s not the neuronal variables that we are
trying to explain. This would be more the domain of basic neuroscience.

As I see it, cognitive neuroscience would be helped a lot if we could
quantify the extent to which a particular neuronal variable can explain
some behavioral variable. That is, we would be computing an effect size
for a behavioral variable, using the neuronal variable as the explanatory
(instead of to-be-explained) variable. Together with Freek van Ede and
Malte Köster, I proposed a method that is able to do precisely this
(Journal of Neurophysiology 108 (9), 2352-2362,
http://jn.physiology.org/content/108/9/2352.short). The crucial
methodological issue here is how to deal with the inherent unreliability
of the measured neuronal variables.

Of course, a basic neuroscientist who is trying to explain to explain the
firing rate of a single neuron may very well be interested in an effect
size for this firing rate which could show, for instance, that this firing
rate is only very weakly determined by the presence/absence of a stimulus.



Best,



Eric Maris





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20130427/a1443cd9/attachment-0002.html>


More information about the fieldtrip mailing list