sensorplot_clusterStatsTFR_2D.m

function [] = sensorplot_clusterStatsTFR_2D(n, sessions)
% select subsets of trials that we'll look at and compare

if ~isdeployed,
    addpath(genpath('~/code/MEG'));
    addpath(genpath('~/code/Tools'));
    addpath('~/Documents/fieldtrip');
    ft_defaults;  % ft_defaults should work in deployed app?
    warning off;
end

% http://www.fieldtriptoolbox.org/faq/matlab_complains_about_a_missing_or_invalid_mex_file_what_should_i_do/#known_issues_with_spm_mex_files_and_workarounds
global ft_default
ft_default.spmversion = 'spm12';
ft_defaults;

% run for n = 2 (visual gamma) and n = 4 (motor beta)
% add in: n = 7 (previous choice)

if ~exist('sessions', 'var'),   sessions = [1 2]; end

% for running on stopos
if ischar(n),             n = str2double(n); end
if ischar(sessions),       sessions = str2double(sessions); end

% define subjects
sj              = 'GAcleantmp';
subjectdata     = subjectspecifics(sj);
[~, conditions] = sensorplot_defineConditions();

for session = sessions,
    
    % use only the locking where we expect the effect
    for d = 1:2,
        locks = {'ref', 'stim', 'resp', 'fb'};
        load(sprintf('%s/%s-S%d_freqbl_%s_%s_allindividuals.mat', ...
            subjectdata.tfrdir, sj, session, locks{conditions(n).timewin{2}}, conditions(n).name{d}));
        fprintf('%s/%s-S%d_freqbl_%s_%s_allindividuals.mat \n', ...
            subjectdata.tfrdir, sj, session, locks{conditions(n).timewin{2}}, conditions(n).name{d});
        
        % only use a small time window
        grandavg = ft_selectdata(struct('avgovertime', 1, 'latency', conditions(n).timewin{1}), grandavg);
        grandavg = rmfield(grandavg, 'cfg'); % keep the structs small
        % grandavg.dimord = 'subj_chan_freq_time';
        grandavg.dimord = 'subj_chan_freq';
        data(d)  = grandavg;
    end
    
    % ==================================================================
    % prepare for group-level stats
    % ==================================================================
    
    load('ctf275_neighb.mat'); % get neighbour struct for clusters
    
    cfg                  = [];
    cfg.channel          = 'MEG'; % use only subset of sensors
    cfg.minnbchan        = 3; % 3, according to Tom (two chans can form weird bridges)
    cfg.neighbours       = neighbours;
    cfg.parameter        = 'powspctrm';
    cfg.method           = 'montecarlo';
    cfg.statistic        = 'ft_statfun_depsamplesT'; % within-subject contrast
    
    cfg.correctm         = 'cluster';
    cfg.clustertail      = 0; % two-tailed
    cfg.clusteralpha     = 0.05;
    cfg.clusterstatistic = 'maxsum'; % sum all t-values in a cluster
    
    cfg.tail             = 0; % two-tailed
    cfg.alpha            = 0.05; % alpha, see below
    cfg.correcttail      = 'prob'; % see http://www.fieldtriptoolbox.org/faq/why_should_i_use_the_cfg.correcttail_option_when_using_statistics_montecarlo
    
    cfg.numrandomization = 10000;
    cfg.randomseed       = sum(100*clock); % keep this for reproducibility
    
    % paired design, within subject
    subj = size(data(1).powspctrm, 1);
    design = zeros(2,2*subj);
    for i = 1:subj
        design(1,i) = i;
    end
    for i = 1:subj
        design(1,subj+i) = i;
    end
    design(2,1:subj)        = 1;
    design(2,subj+1:2*subj) = 2;
    cfg.design   = design;
    cfg.uvar     = 1;
    cfg.ivar     = 2;
    
    stat         = ft_freqstatistics(cfg, data(1), data(2));
    sum(stat.mask(:))
    
    % save with the name of this contrast, also make sure to save the random seed
    save(sprintf('%s/%s-S%d_%s_%s_c%d_freqstats2D.mat', subjectdata.statsdir, ...
        sj, session, conditions(n).name{1}, conditions(n).name{2}, n), 'stat', 'cfg');
end

end