# [FieldTrip] Permutation test with low amount of subjects participating in several sessions

Steinmann, Iris iris.steinmann at med.uni-goettingen.de
Thu Nov 3 17:27:40 CET 2016

```Hello Fieldtrip experts,
currently I'm working on an intracranial dataset with low amount of subjects who participated repeatedly in an experiment with two different conditions.
In detail:

*        LFP data from only three subjects.

*        Each subject participated several times in the same experiment (about 16 sessions per subject).

*        In every session subjects performed around 50 trials of condition A and around 50 trials of condition B
I calculated time-frequency spectra (TFS) for the LFP's. To test if there are significant differences between the TFS(condition A) and TFS(condition B) I want to use the permutation test implemented in fieldtrip. Unfortunately I'm struggling with my little statistic knowledge, because of the low amount of subjects and the high repetition rate for every subject in multiple sessions.
Here is what I have done so far, and it would be great if anyone could tell me if it is correct or totally bullshit.
First I averaged over all trials, so I put one TFS for each condition and session in the statistic. The first row of the design matrix represent the repetition of the single subjects (in this case three) and the second row of the design matrix contains the two conditions A (as 1) and B (as 2).
cfg = [];
cfg.parameter = 'powspctrm';
cfg.numrandomization = 5000;
cfg.method = 'montecarlo';
cfg.correctm = 'fdr';
cfg.alpha = 0.05;
cfg.correcttail = 'prob';
cfg.ivar = 2;
cfg.uvar = 1;
cfg.statistic = 'ft_statfun_depsamplesT';
design = [1 1 1 1 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 1 1 1 1 1 1 1 1 1 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3; 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2];
cfg.design = design;

stat = ft_freqstatistics(cfg, data_A, data_B);