[FieldTrip] Permutation test with cluster based correction on a single sensor with time frequency data
Schoffelen, J.M. (Jan Mathijs)
janmathijs.schoffelen at donders.ru.nl
Fri Dec 20 19:46:31 CET 2024
Hi Anne,
It seems as if you have solved it yourself indeed. I did not notice that the data structure that you pasted in an earlier e-mail indeed was fieldtrip-incompatible, with an incorrect ‘dimord’. I don’t know where this data structure came from, but if it was created using ‘purely’ fieldtrip functions, please do point us to where in the code this happened. Otherwise, I will file this as a ‘user error’, thanks for letting us know, and glad that it now works for you.
Best wishes,
Jan-Mathijs
On 19 Dec 2024, at 16:53, Anne Guérin via fieldtrip <fieldtrip at science.ru.nl> wrote:
Le 19/12/2024 à 14:57, Schoffelen, J.M. (Jan Mathijs) via fieldtrip a écrit :
Hi Anne,
I don’t know what’s going on.
Have you tried cfg.neighbours = {} (i.e. curly rather than square brackets), and/or explicitly defined cfg.connectivity = []; ?
Best wishes,
Jan-Mathijs
Hi Jan-Mathijs,
When I try cfg.neighbours = {}, instead of [], a bug appears earlier at
Unrecognized field name "label".
Error in channelconnectivity (line 91)
[seld] = match_str(chans, neighbours(chan).label);
I solved it with cfg.chanel = {ROI} in order to overlap the field .label of the data structure, and with cfg.neighbours = [];
after this function (channelconnectivity) and go back to
I have for the cfg structure and these 3 fields :
channel: {'V5_R'}
neighbours: []
connectivity: 0
With cfg.neighbours={}; , in the function channelconnectivity, the bug appears at line 91
line 90 for chan=1:length(neighbours)
line 91 [seld] = match_str(chans, neighbours(chan).label);
in fact with cfg.neighbours={} , the variable 'neighbours' is a struct with no field, so length(neighbours)=1 , and the loop begins and it's bug.
So I go back cfg.chanel = {'V5_R'} and cfg.neighbours=[] (cfg.connectivity = [] has no impact), I find the same bug again, at the function 'findcluster'.
But, I think that the problem comes from the expected input data onoff of the function findcluster :
onoff = is a 3D boolean matrix with size N1xN2xN3
three dimensions are expected for chan freq and time, but in my case there are only two dimensions freq and time. So I changed the indut data like this (for example for the first subject) :
ftAllFiles{1}
ans =
struct with fields:
dimord: 'chan_freq_time'
avg: [1×29×401 double]
time: [-0.2000 -0.1990 -0.1980 -0.1970 -0.1960 -0.1950 -0.1940 -0.1930 -0.1920 -0.1910 -0.1900 -0.1890 -0.1880 -0.1870 -0.1860 … ]
freq: [6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34]
label: {'V5_R'}
with a first dimension as a scalar (only one chanel).
It now seems to be working properly. but I must verify the results to be sure.
The moral of the story is that you always have to define a “chan” dimension, even if there's only one chanel, and that it's not theoretically necessary to encode the data in this case.
Thank you.
Best regards.
Anne
On 19 Dec 2024, at 12:18, Anne Guérin via fieldtrip <fieldtrip at science.ru.nl><mailto:fieldtrip at science.ru.nl> wrote:
Hi,
I would like to use the FT permutations test with cluster based correction on 2D data freq x time with a within subject design. Data are obtained after sources reconstruction and are for example related to a particular ROI (e.g. V5_R). The first dimension is a frequency dimension but not the oscillatory brain frequency. This dimension is linked to a temporal frequency of the input visual stimulus. the second dimension is time (sRate = 1kHz) and data are timelocked ERP for different values of this temporale frequency (by step of 1Hz). The data therefore has the same structure as if it had been obtained by time-frequency analysis on a particular electrode.
For example, here's the data for one participant:
ftAllFiles{1}
ans =
struct with fields:
dimord: 'freq_time'
avg: [29×401 double]
time: [-0.2000 -0.1990 -0.1980 -0.1970 -0.1960 -0.1950 -0.1940 -0.1930 -0.1920 -0.1910 -0.1900 -0.1890 -0.1880 -0.1870 -0.1860 … ]
freq: [6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34]
label: {'V5_R'}
There are 29 participants for each conditions. ftAllFiles have 2x29 cells, the first 29 correspond to the first condition and the last 29 to the second condition.
Here is the configuration :
>> statcfg
statcfg =
struct with fields:
frequency: 'all'
parameter: 'avg'
avgoverfreq: 'no'
avgoverchan: 'no'
latency: 'all'
method: 'montecarlo'
statistic: 'ft_statfun_depsamplesT'
correctm: 'cluster'
clusteralpha: 0.0500
clusterstatistic: 'maxsum'
minnbchan: 0
tail: 1
clustertail: 1
alpha: 0.0500
numrandomization: 100000
ivar: 1
uvar: 2
design: [2×58 double]
channel: 'all'
neighbours: []
avgovertime: 'no'
When I run ftStat = ft_freqstatistics(statcfg, ftAllFiles{:}); , the program stops at the step 'findcluster':
posclusobs = findcluster(tmp, connmat, cfg.minnbchan);
In fact tmp is 2D boolean table (29 frequency samples x 401 time samples), connmat is 4x4 matrix with 0 and cfg.minnbchan=0. So bug with findcluster. See below.
When I plot tmp (image(tmp)), the result doesn't look ridiculous.
How to have a good configuration of neighborhood for this single-sensor time-frequency permutation test ?
Thank you for your help.
Best regards.
Anne
ps :
>> ftStat = ft_freqstatistics(statcfg, ftAllFiles{:});
the call to "ft_selectdata" took 0 seconds
using "ft_statistics_montecarlo" for the statistical testing
using "ft_statfun_depsamplesT" for the single-sample statistics
constructing randomized design
total number of measurements = 58
total number of variables = 2
number of independent variables = 1
number of unit variables = 1
number of within-cell variables = 0
number of control variables = 0
using a permutation resampling approach
repeated measurement in variable 2 over 29 levels
number of repeated measurements in each level is 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
the maximum number of unique permutations is 536870912
generated 100000 random permutations
computing a parametric threshold for clustering
computing statistic
estimated time per randomization is 0.00 seconds
computing statistic 100000 from 100000
Error using findcluster
invalid dimension of spatdimneighbstructmat
Error in clusterstat (line 214)
posclusobs = findcluster(tmp, connmat, cfg.minnbchan);
Error in ft_statistics_montecarlo (line 364)
[stat, cfg] = clusterstat(cfg, statrand, statobs);
Error in ft_freqstatistics (line 194)
[stat, cfg] = statmethod(cfg, dat, design);
--
Anne Guérin-Dugué
PR Emérite Université Grenoble Alpes (UGA)
GIPSA-lab / PSD / Equipe ViBS
Bureau B146
Site Ampère
11 rue des Mathématiques
BP 46
F - 38042 GRENOBLE
tel : +33 (0)4 76 57 43 73
mel :anne.guerin at gipsa-lab.grenoble-inp.fr<mailto:anne.guerin at gipsa-lab.grenoble-inp.fr>
_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.science.ru.nl%2Fmailman%2Flistinfo%2Ffieldtrip&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920048561%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=At%2F5yqnrvxE99SiJ5RTk%2FSO7GP5BAGQ3YH9r7WPAQYQ%3D&reserved=0>
https://doi.org/10.1371/journal.pcbi.1002202<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoi.org%2F10.1371%2Fjournal.pcbi.1002202&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920077520%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=O4lUpVLSrmqvmUE8vszk1pDd5PmCxq5jYrRuh743VJs%3D&reserved=0>
_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.science.ru.nl%2Fmailman%2Flistinfo%2Ffieldtrip&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920098126%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=KQvoYgLU%2FDyoqFy5lporUa%2F%2BVr3d2NbZs35h34OPVDY%3D&reserved=0>
https://doi.org/10.1371/journal.pcbi.1002202<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoi.org%2F10.1371%2Fjournal.pcbi.1002202&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920118395%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=HK60qVI3rApcqoEcqM0Hls2W6WDefRWO7UebPjfXzKk%3D&reserved=0>
--
Anne Guérin-Dugué
PR Emérite Université Grenoble Alpes (UGA)
GIPSA-lab / PSD / Equipe ViBS
Bureau B146
Site Ampère
11 rue des Mathématiques
BP 46
F - 38042 GRENOBLE
tel : +33 (0)4 76 57 43 73
mel :anne.guerin at gipsa-lab.grenoble-inp.fr<mailto:anne.guerin at gipsa-lab.grenoble-inp.fr>
_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmailman.science.ru.nl%2Fmailman%2Flistinfo%2Ffieldtrip&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920134097%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=4AxcIo94hYp%2BIY6Emlo0SIBzB%2BwrMB7rPPMgHCBipwo%3D&reserved=0>
https://doi.org/10.1371/journal.pcbi.1002202<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoi.org%2F10.1371%2Fjournal.pcbi.1002202&data=05%7C02%7Cfieldtrip%40science.ru.nl%7C1a5c19cd43ae41ccbb5608dd21269e9e%7C084578d9400d4a5aa7c7e76ca47af400%7C1%7C0%7C638703171920147734%7CUnknown%7CTWFpbGZsb3d8eyJFbXB0eU1hcGkiOnRydWUsIlYiOiIwLjAuMDAwMCIsIlAiOiJXaW4zMiIsIkFOIjoiTWFpbCIsIldUIjoyfQ%3D%3D%7C0%7C%7C%7C&sdata=4jZ1KoVsTY5NZv%2FL%2FRArnYbU47rd6DwlExDm00PYYQQ%3D&reserved=0>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20241220/41db2e3e/attachment.htm>
More information about the fieldtrip
mailing list