[Eeglablist] Codes for cluster correction are needed
Dr Cyril, Pernet
wamcyril at gmail.com
Thu Aug 4 21:25:20 PDT 2022
Hi Olivia
>
> Hi Cyril,
>
> Thanks a lot!
>
> I think you are right. It would not be reasonable if this function
> without the inputs of channel locations could finish spatial cluster
> correction.
>
> It’s really difficult to mine some user-friendly EEGLAB-based codes
> online for the spatial cluster correction on peak statistics.
>
because it doesn't exist! peaks makes sense in the context of neighbours
(like my code, looking for mountains to threshold) but because your
matrix are extracted peak values that info is gone - I do think your
only solution is FDR -- you can use this
https://urldefense.com/v3/__https://github.com/LIMO-EEG-Toolbox/limo_tools/blob/master/limo_FDR.m__;!!Mih3wA!E_U3iY-VznFXsp5FOyHSCEDv8aQBT3C86SFsHUENG13WwMqXxPgo4tN4DWP41zfrv_nDtWX2k9dMXOLsvp4Clg$ ,
which clearly doesn't care about channels (ie input your matrix M as M(:))
Cyril
> ------- Original Message -------
> Dr Cyril, Pernet wamcyril at gmail.com wrote:
>
>
> I don't see how that code would work ?? Olivia's data are
> SubjectsConditionsChannels for peak values
>
> just switch dim to please the code Channels * Subjects say cond1 vs
> cond2, the difference cannnot create good clster over space because
> channel is vectorized and thus can be organized in any way possible
>
> sorry Olivia - I'm pretty sure it won't work ...
> cyril
>
>
> ------- Original Message -------
>
> On 27/07/2022 16.16, Olivia Yang via eeglablist wrote:
>
> Hi Makoto,
>
>
> Thank you very much!
>
>
> Sorry to bother you, but I have the following questions:
>
>
> 1. Are there demo codes for this solution to do two-factor repeated ANOAV?
>
> If not, am I correct with my try:
>
>
> %%%%%%%%%%%%%%%%%%%%%%%%%%%%
>
>
> for c=1:4
>
> input{i}=squeeze(PeakValue(c,:,:));% PeakValue is a three dimensional
> matrix (Experimental conditions × Channels × Subjects) of peak values
> picked out from N170；
>
> % Four experimental conditions belong to two factors：
>
> %input{1} is the date for level 1 of factor A and level 1 of factor B
>
> %input{2}, level 2 of factor A and level 1 of factor B
>
> %input{3}, level 1 of factor A and level 1 of factor B
>
> %input{4}, level 2 of factor A and level 2 of factor B
>
> end
>
> repeatedMeasuresFlag= 3;% random-effect test
>
> pValForPreselection=0.05;% Is the p-value before or after cluster
> correction needed here? Is there a recommended setting for this
> parameter?
>
> numIterations=10000;
>
> [mask, tScore, pValue] = clusterLevelPermutationTest2x2(input{1},
> input{2), input{3}, input{4}, repeatedMeasuresFlag,
> pValForPreselection, numIterations);
>
>
> %%%%%%% %%%%%%%%%%%%%%%%%%%%%%%
>
>
> 2. About the outputs of this function, I don’t know which parts
> indicate the cluster correction results of the interaction effect or
> the main effects of ANOAV analysis.
>
>
> 3. Are not channel locations needed as inputs? That makes me wonder
> how this function conducts multiple comparison corrections across
> space without channel location information.
>
>
> Best regards,
>
> Olivia
>
>
>
> ------- Original Message -------
> Makoto Miyakoshi via eeglablist eeglablist at sccn.ucsd.edu wrote:
>
> Dear Olivia,
>
> In case you be interested in my solution, see the
> following code for your
> info.
> https://urldefense.com/v3/https://github.com/sccn/groupSIFT/blob/master/clusterLevelPermutationTest.m;!!Mih3wA!FcwOZGypevvYbX4NXfyvIOJRXVzzgjj1amwOZGA8N4-LyVn9QVPYjt9Qj-zN-qeE36ZGpxTN1zdSmayiorkXQOSfAbQ$
>
> Makoto
>
> On Thu, Jul 21, 2022 at 11:35 AM Dr Cyril, Pernet via
> eeglablist <
> eeglablist at sccn.ucsd.edu> wrote:
>
> Hi Olivia,
>
> You can in principle do clustering under 1 dimension
> only using
>
> https://urldefense.com/v3/https://github.com/LIMO-EEG-Toolbox/limo_tools/blob/master/limo_clustering.m;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2UUzu6Ayw$
>
> - but you will have to generate the bootstrap or
> permutation data under
> the null. I am really not sure that this would be less
> conservative than
> FDR in such a case though.
>
> Cyril
>
> Hi Cyril，
>
> Thank you very much for your kind reply with
> detailed explanations!
>
> However, FDR correction is too conservative in my
> case. So, I’d like to
> use alternative correction methods that can take
> space information into
> consideration.
>
> I thought it would be easy to finish cluster
> corrections for peak
> amplitude statistics in EEGLAB. But it looks like not.
>
> Best wishes,
> Olivia
>
> ------- Original Message -------
> On Wednesday, July 20th, 2022 at PM 3:18, Dr
> Cyril, Pernet <
> wamcyril at gmail.com> wrote:
>
> Hi Olivia
>
> clustering requires some continuity over space
> and/or time - here you
> have a 30 (Subjects) × 4 (Conditions in my
> experiment) × 66 (Channels)
> matrix of peak values so could cluster over
> space but I would really not
> recommend doing that as these a 'peaks' so
> really different measurements
> at each channels - in your case I'd test
> whatever differences you are
> interested in and apply an FDR correction.
>
> more info from a good old EEGLAB lecture
> https://urldefense.com/v3/https://figshare.com/articles/presentation/Statistical_analysis_of_EEG_data_Hierarchical_modelling_and_multiple_comparisons_correction/4233977;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2Wvu8_cjQ$
>
> slides 2 and 3 gives you an idea why I'd not
> recommend your approach
> above and slides 24+ explains multiple
> comparison corrections and among
> those procedure, clustering.
>
> reference article
> https://urldefense.com/v3/https://www.sciencedirect.com/science/article/pii/S0165027014002878;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2WxKJjPFg$
>
> cyril
>
> Dear EEGLAB geniuses,
>
> Do you know whether there are codes for
> performing cluster corrections?
>
> After I finished the preprocessing
> analysis in EEGLAB, I picked out
> peak values of N170 from average waveforms
> through custom Matlab codes. And
> a 30 (Subjects) × 4 (Conditions in my
> experiment) × 66 (Channels) matrix of
> peak values was generated. Afterwards, I
> conducted repeated measure ANOVA
> and got statistical values of each channel.
>
> Now I need to perform multiple comparison
> corrections (cluster
> correction) for the comparison of peak
> value difference among the four
> conditions. Are there codes or more
> effortless ways that can help me?
>
> Looking forward to your reply!
>
> Cheers!
>
> Best regards,
>
> Olivia
> _______________________________________________
> 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
> --
> Dr Cyril Pernet, PhD, OHBM fellow, SSI fellow
> Neurobiology Research Unit,
> Building 8057, Blegdamsvej 9
> Copenhagen University Hospital, Rigshospitalet
> DK-2100 Copenhagen, Denmark
>
> wamcyril at gmail.com
> https://urldefense.com/v3/https://cpernet.github.io/;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2UdSb0_VA$
>
> https://urldefense.com/v3/https://orcid.org/0000-0003-4010-4632;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2Vu9feMUA$
>
> --
> Dr Cyril Pernet, PhD, OHBM fellow, SSI fellow
> Neurobiology Research Unit,
> Building 8057, Blegdamsvej 9
> Copenhagen University Hospital, Rigshospitalet
> DK-2100 Copenhagen, Denmark
>
> wamcyril at gmail.com
>
> https://urldefense.com/v3/https://cpernet.github.io/;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2UdSb0_VA$
>
> https://urldefense.com/v3/https://orcid.org/0000-0003-4010-4632;!!Mih3wA!BrNn2j4iKMeFLS6PS8IC6f7nRqK9ye2CG39s_8-u6R209tyg6NpxXExCVvR9z9g0EU9_RjUBrWmEr2Vu9feMUA$
>
> _______________________________________________
> 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
> _______________________________________________
> 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
>
> --
> Dr Cyril Pernet, PhD, OHBM fellow, SSI fellow
> Neurobiology Research Unit,
> Building 8057, Blegdamsvej 9
> Copenhagen University Hospital, Rigshospitalet
> DK-2100 Copenhagen, Denmark
>
> wamcyril at gmail.com
> https://urldefense.com/v3/__https://cpernet.github.io/__;!!Mih3wA!E_U3iY-VznFXsp5FOyHSCEDv8aQBT3C86SFsHUENG13WwMqXxPgo4tN4DWP41zfrv_nDtWX2k9dMXOIo1Fjgqw$
> https://urldefense.com/v3/__https://orcid.org/0000-0003-4010-4632__;!!Mih3wA!E_U3iY-VznFXsp5FOyHSCEDv8aQBT3C86SFsHUENG13WwMqXxPgo4tN4DWP41zfrv_nDtWX2k9dMXOLAnP7ynA$
>
--
Dr Cyril Pernet, PhD, OHBM fellow, SSI fellow
Neurobiology Research Unit,
Building 8057, Blegdamsvej 9
Copenhagen University Hospital, Rigshospitalet
DK-2100 Copenhagen, Denmark
wamcyril at gmail.com
https://urldefense.com/v3/__https://cpernet.github.io/__;!!Mih3wA!E_U3iY-VznFXsp5FOyHSCEDv8aQBT3C86SFsHUENG13WwMqXxPgo4tN4DWP41zfrv_nDtWX2k9dMXOIo1Fjgqw$
https://urldefense.com/v3/__https://orcid.org/0000-0003-4010-4632__;!!Mih3wA!E_U3iY-VznFXsp5FOyHSCEDv8aQBT3C86SFsHUENG13WwMqXxPgo4tN4DWP41zfrv_nDtWX2k9dMXOLAnP7ynA$
More information about the eeglablist
mailing list