Bug789

This shows the difference before and after the bug fix. The top row is before the bug fix and the bottom row after the bug fix. As we can the difference is small for this type of data.

p<0.05	p<0.05 (FDR)

Numerical threshold for positive power at p<0.05 at each frequency (before bug fix)

  1.0e+04 *
   3.1245    2.9186    2.6326    2.2681    1.7952    1.4107    1.0380    0.7739    0.5857    0.4723
   0.4106    0.3864    0.3674    0.3570    0.3357    0.3223    0.2993    0.2826    0.2622    0.2436
   0.2262    0.2047    0.1903    0.1746    0.1617    0.1510    0.1384    0.1261    0.1153    0.1054
   0.0982    0.0903    0.0865    0.0822    0.0788    0.0750    0.0728    0.0686    0.0643    0.0596
   0.0549    0.0510    0.0466    0.0424    0.0379

Numerical threshold for positive power at p<0.05 at each frequency (after bug fix)

  1.0e+04 *
   3.4005    3.0917    2.7952    2.4521    1.9075    1.5034    1.1044    0.8410    0.6374    0.5087
   0.4385    0.4168    0.3998    0.3915    0.3639    0.3514    0.3235    0.3002    0.2780    0.2594
   0.2388    0.2203    0.2083    0.1850    0.1761    0.1646    0.1464    0.1355    0.1217    0.1134
   0.1054    0.0959    0.0916    0.0901    0.0843    0.0790    0.0778    0.0730    0.0686    0.0630
   0.0584    0.0538    0.0500    0.0461    0.0407

The threshold is higher after bug fix. Despite the fact that we barely see any difference between the time frequency decompositions, the upper array and pictures correspond to p < 0.2. Note that only ERD (Event-Related Desynchronization) and not positive power changes are affected.

The figures above were obtained using the tutorial dataset and the following Matlab commands. For non-FDR:

 figure; pop_newtimef( EEG, 0, 1, [-1000  1992], [3         0.5] , 'topovec', ...
    EEG.icawinv(:,1), 'elocs', EEG.chanlocs, 'chaninfo', EEG.chaninfo, ...
    'baseline',[0], 'alpha',0.05, 'plotitc' , 'off', 'plotphase', 'off', 'padratio', 4);

Using FDR:

 figure; pop_newtimef( EEG, 0, 1, [-1000  1992], [3         0.5] , 'topovec', ...
    EEG.icawinv(:,1), 'elocs', EEG.chanlocs, 'chaninfo', EEG.chaninfo, ...
    'baseline',[0], 'alpha',0.05, 'plotitc' , 'off', 'plotphase', 'off', ...
    'padratio', 4, 'mcorrect', 'fdr', 'naccu', 2000);