The Neurophysiological Biomarker Toolbox (NBT)


Table of Contents

Compute Statistics

The NBT GUI has built-in tools for statistical analysis of biomarkers (List of statistical tests).

For the Demoset, we intend to detect statistical differences between Eyes Closed Rest condition and Eyes Open Rest condition. Recordings are from the same subjects who underwent both conditions. For this reason we will perform a paired-test like paired t-test, which implies that the timeseries are normally distributed, or a Wilcoxon signed-rank test, if we prefer a distribution free test.

Open NBT GUI.

  • Go to Biomarker statistics|Statistics GUI. The following window for group settings will appear.

  • Click the button Define new groups, and type 2 in the popup window asking how many groups (ECR and EOR) you want to consider.
  1. Select the folder where the analysis files related to the first group are stored (for the demoset the ECR and EOR files are saved in the same folder, make sure the info files are also stored in the same folder). Once the folder has been selected NBT will inspect each file contained in the folder, distinguishing for conditions, project, subjectID, date, age and gender of the subject if present (age and gender are not included in this demoset).
  2. The following interface will appear and you will be able to select the first group. Select ECR1 in the 'conditions' listbox and assign a name to the group (i.e., EyesClosedRest), then click OK and wait for NBT to sort the files relative to the ECR condition.
  3. Repeat steps 1 and 2 for the second group (this time select EOR1 in the 'conditions' listbox and assign a new name to the second group, i.e., EyesOpenRest)

The Main statistics GUI now appears (see below)

  • First select the statistical test you want to calculate in the “Select Test” panel. Select “Students paired t-test”
  • Then proceed by selecting the biomarkers for the statistics in the “Select Biomarker(s)” panel. Select the DFA biomarkers and the Amplitude.Channels biomarkers (see image).

  • Click the button Select Channels and Regions in the statistics settings window. Here you can specify the channels you are interested in and/or the regions you want to evaluate. Select All channels and load standard regions file clicking Load Regions (the standard regions file is contained in the Statistics folder in the NBT toolbox folder). Click Submit.

  • Select the two groups you want to run the test on in the “Select Group(s)” panel, and click Run test.

A Student's paired t-test for the two conditions for each biomarker will be computed. The outcome of the computation is a map showing p-values for each biomarker (vertical axis) and each channel (horizontal axis). P-values < 0.05 indicate significant statistical differences between the two conditions. As expected the biomarker that shows high statistical difference is the power (or amplitude) of the alpha oscillations (8-13Hz).

Now, if you right click on a biomarker name (e.g. amplitude 8 13Hz.Channels) on the left of the map you will be able to inspect the biomarkers values in the two conditions across the scalp (via topoplot) and the results for the statistical test performed on the difference between the two conditions (via errorbars and confidence intervals):

With a right click on a specific pixel on the map (e.g. raw of amplitude 8 13Hz.Channels column = channel 74) you can have an overview of the biomarker values in a specific channel for all the subjects (identified by their ID number), together with a boxplot, estimating mean values and confidence interval in the two conditions. This plot will allow you also to visualize the presence of potential outlier subjects in the pool of your data:

Go back to the main statistics window, try to select Regions instead of Channels in the “Select Channels or Regions” panel, and Run test again. A new p-values map will appear, in this case the biomarkers will be evaluated for regions instead of channels. And you can clearly see in which brain region and for which biomarker the difference between the two conditions is higher.

Again you can right click on the name of a biomarker or on one pixel of the map to zoom-in into the analysis.

With this demoset we showed how the analysis of EEG signals can be both straight forward and also enriched by multiple biomarkers computation.

Compare the results of your analysis with the analysis files provided in the demoset:

To make sure that the cleaning of your data didn't affect the biomarker computation and thus the statistics outcome, you can try to compare your analysis files with the given analysis files (in the cleaned demoset folder) using the Statistics GUI.

You can define four groups: DemosetECR and DemosetEOR, MyDemosetECR and MyDemosetEOR, the files related to the first two groups are stored in the cleaned demoset folder, the files related to the second two groups are the ones that you created by following the steps of data cleaning and biomarker computation.

Select specific biomarkers and channels or regions and then run a paired t-test between DemosetECR vs MyDemosetECR, first, and between DemosetEOR vs MyDemosetEOR, next. If no statistical differences are found among biomarkers between your data and the demoset, congratulations! It means that you did a good job and you obtained the same results as in the demoset.

Go back to the Tutorials

tutorial/compute_statistics.txt · Last modified: 2014/02/05 14:07 by Simon-Shlomo Poil
The NBTwiki platform - version 2.8 - 9 May 2013
Copyright (C) 2008-2015