<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2900.2769" name=GENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=#ffffff>
<DIV>Hi Marco,</DIV>
<DIV> </DIV>
<DIV> </DIV>
<BLOCKQUOTE dir=ltr style="MARGIN-RIGHT: 0px">
<DIV>thank you for your accurate responses. I fully understand from your
arguments that temporally zooming on clusters is definitely wrong. Still, I
wonder whether and how it is possible to use cluster randomization analysis
cases in which it is difficult to formulate a precise hypothesis about when to
expect an effect (for example, in infants), or cases in which an
unexpected effect arises from a t-test. Do you think it would be correct
to slide a relatively large (width of 200ms? 400ms? to be chosen a priori of
course) window through the epochs and compute cluster randomization
analysis for each latency to explore dubious significant t-test clusters?
</DIV></BLOCKQUOTE>
<DIV> </DIV>
<DIV>If you have no hypothesis about where to expect an effect, you should use
the complete latency window in which it may occur. Of course, this will reduce
the sensitivity (statistical power) of your test (in comparison with the
situation in which you do know when the effect can occur). As a rule,
prior knowledge increases sensitivity.</DIV>
<DIV> </DIV>
<DIV> </DIV>
<BLOCKQUOTE
style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<BLOCKQUOTE dir=ltr style="MARGIN-RIGHT: 0px">
<DIV>Another related question: I computed a post-hoc non kosher tuning of
the window around the most significative cluster in my data, and I saw
that it is significative (p<0.05) if the window edges exceed
of about 50 ms the cluster edges (since the cluster is about 70 ms
long, the whole window is about 170 ms long); but if I take longer windows,
the p-value increases quite rapidly (I'm running at least 500 random
draws for each window, and checking that the result does not depend on the
number of draws). Do you have such instabilities in your data or should
I think that the effect relative to my cluster is definitely too weak? Or
maybe my data are not clean enough? </DIV></BLOCKQUOTE>
<DIV> </DIV>
<DIV>This phenomenon is not an instability, it is what I would expect. Imagine
your trials are 10 seconds long and there is an effect in the latency
window between 1.3 and 1.35 seconds (i.e., less than 1 percent of trial
length). If you ask clusterrandanalysis to compare the conditions over the
complete trial length, it may very well miss the effect in the window between
1.3 and 1.35 seconds, because it has to use a large critical value in
order to control for false positives in the time window where there is no
effect (i.e., 99 percent of the 10 second trial). </DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV>greetings,</DIV>
<DIV> </DIV>
<DIV>Eric Maris</DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV><SPAN class=gmail_quote>On 10/28/05, <B class=gmail_sendername>Eric
Maris</B> <<A href="mailto:maris@nici.ru.nl">maris@nici.ru.nl</A>>
wrote:</SPAN>
<BLOCKQUOTE class=gmail_quote
style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">Dear
Marco,<BR><BR><BR>> The procedure I am following now is a sort of
two-steps method: in the<BR>> first place, I choose a wide time interval
and a low minimum number of <BR>> channels. I end up with many clusters
that are far from being<BR>> significative. I then shorten the time
interval to include just one<BR>> cluster (starting from the most
significant one), and increase the minimum <BR>> number of channels, and
run the analysis again. In this case, I eventually<BR>> got a
significative cluster where I was expecting it from a simple<BR>>
observation of the t-test. Do you think this procedure is right or am I
<BR>> doing something wrong? Is it correct to temporally focus on a
cluster to<BR>> check its significance?<BR><BR><BR>Clusterrandanalysis
only controls the false alarm (type I error) rate if you<BR>choose the
"tuning parameters" (latency interval, channel subset, the
<BR>minnbchan-parameter; and if you use on TFRs, also the frequency
interval)<BR>independent of the data. Instead, if you play around with these
tuning<BR>parameters until you find a cluster whose p-value exceeds the
critical <BR>alpha-level, you are not controlling the false alarm rate. In
this case, the<BR>chosen tuning parameters depend on the data.<BR><BR>An
extreme example illustrates this even better. Assume you
calculate<BR>T-statistics for all (channel, time point)-pairs and you select
the pair <BR>with the largest T-statistic. Then, you select the latency
interval that<BR>only contains this time point and the channel subset that
only contains this<BR>channel. With these tuning parameters, you reduce your
data to a single cell <BR>in the spatiotemporal matrix, and
clusterrrandanalysis will produce a<BR>p-value that is very close to the
p-value of a T-test. Since you have<BR>selected this (channel, time
point)-pair on the basis of its T-statistic, <BR>this p-value is strongly
biased.<BR><BR><BR>> Another couple of questions:<BR>> 1) Minnbchan. I
understood it is the minimum number of significative<BR>> neighbor
(channel,time) points for a (channel,time) point to enter a
<BR>> cluster, no matter if adjacency is more in channel space or
time<BR>> direction. Am I right? Since time and channel space are quite
different<BR>> dimension, would it be better to set a minimum channel
number separately <BR>> for the two?<BR><BR>Minnbchan should also be
chosen independent of the data. I introduced this<BR>tuning parameter
because it turned out that in 3-dimensional analyses on<BR>TFRs (involving
the dimensions time, space ( i.e., sensors) and frequency),<BR>sometimes a
cluster appeared that consisted of two or more 3-dimensional<BR>"blobs" that
were connected by a single (channel, time, frequency)-element.<BR>From a
physiological perspective, such a cluster does not make sense. To <BR>remove
these physiologically implausible (and therefore probably
random)<BR>connections, I introduced the minnbchan parameter. Because of
this<BR>physiological rationale, I apply the minimum number criterium to
the<BR>spatial, and not to the temporal dimension. Short-lived phenomena are
very<BR>well possible from a physiological perspective, whereas effects at
spatially<BR>isolated sensors are not.<BR><BR><BR>> 2) Maybe because my
data are average-referenced, I often end up with a <BR>> positive and
negative cluster emerging almost at the same time. Have you<BR>> thought
about any way to include the search of dipole-like configurations?<BR><BR>I
have not thought about it, but it certainly makes sense to incorporate
<BR>biophysical constraints (such dipolar patterns) in the test
statistic.<BR><BR>One should be aware of the fact that different hypotheses
are tested before<BR>and after rereferencing. This is physical and not a
statistical issue. As <BR>you most certainly know, EEG-signals are potential
DIFFERENCES and therefore<BR>the underlying physiological events that are
measured by EEG depend on the<BR>reference channel(s). If the experimental
manipulation affects the current <BR>reference channel, then rereferencing
to another channel (or set of<BR>channels) that is not affected by the
experimental manipulation makes a<BR>difference for the result of the
statistical test.<BR><BR><BR>greetings, <BR><BR>Eric
Maris<BR></BLOCKQUOTE></DIV><BR><BR clear=all><BR>-- <BR>Marco Buiatti - Post
Doc<BR><BR>**************************************************************<BR>Cognitive
Neuroimaging Unit - INSERM U562<BR>Service Hospitalier Frederic
Joliot, CEA/DRM/DSV <BR>4 Place du general Leclerc, 91401 Orsay cedex,
France<BR>Telephone: +33 1 69 86 77 65 Fax: +33 1 69 86
78 16<BR>E-mail: <A
href="mailto:marco.buiatti@gmail.com">marco.buiatti@gmail.com</A> Web:
<A
href="http://www.unicog.org">www.unicog.org</A><BR>***************************************************************
</BLOCKQUOTE></BODY></HTML>