Statistics, visualization, and data export

Statistics

The simple statistics module is limited to pairwise significance tests between any defined data configurations (DCs). For setting up DCs see here. These configurations can include any combination of selected conditions, groups, or sessions.

To perform statistical contrasts and export connectivity data, a pair of DCs needs to be set with the con.contrast2test parameter. The frequency range in Hz for which connectivity will be estimated is defined with the con.freqRange and the associated name (visible in the export files) by the con.subTitle variable. The significance threshold (used for visualization purposes) can be applied with the con.pThresh. To choose between 'within' or 'between' comparisons, con.testType needs to be set, while one- or two-tailed tests are defined by the con.oneTailed option. As connectivity data tends to have skewed distribution, it is recommended to set the con.isParametric to 0, which will result in performing the Wilcoxon signed rank test instead of its parametric version. Moreover, to remove outlier values that can still be present in connectivity data, the IQR rejection methods can be applied before statistical testing by setting the con.iqr (with a recommended value around 3). By default, statistics are only performed for leakage-corrected data, as uncorrected results are usually contaminated with many spurious effects. To see the statistics for the uncorrected data set, set con.unc parameter to 1.

For user convenience, some of the above parameters can include multiple values:.

• con.contrast2test,

• con.freqRange and con.subTitle (both have to match),

• con.pThresh,

• con.iqr

If so, these parameters will be sequentially read, and multiple analyses will be performed within a single statistical run. In the example below, two contrasts will be checked, each one in three frequency bands, and results will be reported using two levels of statistical threshold. Overall, this will produce 2 x 3 x 2 outputs.

con.contrast2test = {[1,2], [3,4]};
con.subTitle  = {'theta, 'alpha', 'beta'};
con.freqRange = {[3 7], [8 12], [15 30]};
con.pThresh   = {0.01, 0.001};
con.iqr = {3};           % remove extremes before stat
con.oneTailed = true;    % divides p-values from 2-tailed test by 2
con.testType = 'within'; % must be within or between

Data export

Data export is performed together with running statistical comparisons (as defined by the con.contrast2test). Group connectivity data are then saved into the xls spreadsheet together with participant Id, group, condition, and session data. By default, all combinations of connectivity directions are included. In the spreadsheet header, all relevant information is provided, including from/to ROI names.

Exported files include:

• xls spreadsheets with connectivity data (if con.iqr parameter is set, values exceeding the threshold will turn into missing data),

• significance matrices of nroi x nroi size, showing thresholded p-levels for all connections between defined ROIs.

• boxplots for directions that passed the significance test, depicting the distributions of DTF values at both compared DCs.

p-level significance matrix showing three significant effects for the inco-cor-PA-post vs inco-cor-SB-post contrast. Positive p-level numbers mean higher DTF values for the first comparison element (here: inco-cor-PA-post) compared to the second element (inco-cor-SB-post). The directions should be read from columns (x) to rows (y).

Boxplots of the significant effects

For more advanced statistical modeling, exported DTF values can be analyzed using external statistical packages.

Connectivity visualization

Significant effects are visualized on the 3D head images. Both vector images in Matlab fig format, which allow for custom rotation, and static jpg top-view figures are saved.

Visualization of the connectivity from the significance matrix above.