Last release: 1.0.0beta
(May 17, 2013)

Previous topic

Subnetwork analysis configuration

Next topic

Extra features configurations

This Page

Local analysis configuration

The local analysis can focus either on the nodes (Local node analysis) or on the edges (Local edge analysis)

Local node analysis

../_images/node_analysis.jpg

Study directory

Specify the directory where “Group1” and “Group2” are located

Data separator

If the data is in csv format, select the character that separates columns in the file. For matlab ”.mat” files, this setting is ignored

Regress data

Select if you want to regress information from the data before performing the statistical analysis. Specify a file where the regressors are stored. Note that only csv files are supported. This file should contain one line per subject (Group1 + Group2), one column with the subject filename (only the file name, eg: file_for_Subj1.csv) and one extra column per regressor The Connectome Analyzer will select the best model that fits the edge measures with the regressors, and remove the contribution of the regressors from the node measures. This is done separately for each node measure

Node measures

Select the set of measures you want to test. They can be:
  • Degree: computes the nodal degree
  • Betweenness: computes the nodal betweenness
  • Closeness: computes the nodal closeness
  • Diameter: computes the nodal diameter

Relaxation parameters

Decomposition method
  • Automatic: decomposes the network subnetworks using the algorithm of your choice (walktrap, leading eigenvector or spinglass).
  • File: provide a file containing the decomposition of the nodes into sub-regions. The file should have 2 columns and as much rows as nodes. The first column should contain the node label, ranging from 1 to the maximum number of nodes. The second column should contain the sub-region label to which the corresponding node belongs. Sub-regions labels should range from 1 to the number of sub-regions without skipping numbers
Correction
Select the method used for the correction for multiple testing
Test type
Define if the test should be parametric (Student t-test) or non-parametric (Wilcoxon test)
Pairing
Define if the data is paired or unpaired. If the data is paired, make sure that the paired files have the same alphanumerical rank in the two folders, otherwise R will load matrices that do not correspond
Alternative
Define the H1 hypothesis (Group1 != Group2, Group1 > Group2 or Group1 < Group2)
Alpha1
Threshold to detect significantly different sub-networks
Alpha2
Threshold bellow which the null hypothesis Group1 == Group2 can be rejected in favor of H1 after correction for multiple testing
Strong
Factor by which the p-values of non-significatively different sub-networks should be multiplied in the RMIO procedure

Local edge analysis

../_images/connection_analysis.jpg

Directories

The local edge analysis can perform uni or multivariate analyses. Add one or several study directories, each one containing a “Group1” and a “Group2” folder

Data separator

If the data is in csv format, select the character that separates columns in the file. For matlab ”.mat” files, this setting is ignored

Mask data

Select if you want to apply a mask to the data before running the analysis. The mask should be a file containing a matrix with the same dimensions as the ones for all subjects. The Connectome Analyzer will put to zero all the values from the connectivity matrices that have a zero value in the mask

Regress data

Select if you want to regress information from the data before performing the statistical analysis. Specify a file where the regressors are stored. Note that only csv files are supported. This file should contain one line per subject (Group1 + Group2), one column with the subject filename (only the file name, eg: file_for_Subj1.csv) and one extra column per regressor The Connectome Analyzer will select the best model that fits the values of each connection with the regressors, and remove the contribution of the regressors from the connections. This is done separately for each connection

Subnetwork summary function

  • MeanEdgeWeights: computes the mean of the connection values that are above zero
  • Mean: computes the mean of all the connections belonging to the subnetwork (takes into account zeroes and negative values)

Relaxation parameters

Decomposition method
  • Automatic: decomposes the network subnetworks using the algorithm of your choice (walktrap, leading eigenvector or spinglass).
  • File: provide a file containing the decomposition of the nodes into sub-regions. The file should have 2 columns and as much rows as nodes. The first column should contain the node label, ranging from 1 to the maximum number of nodes. The second column should contain the sub-region label to which the corresponding node belongs. Sub-regions labels should range from 1 to the number of sub-regions without skipping numbers
Correction
Select the method used for the correction for multiple testing
Test type
Define if the test should be parametric or non-parametric (Student t-test or Wilcoxon test for univariate analyses and parametric Hotelling test or permutation Hotelling test for multivariate analyses). If the analysis is multivariate and non-parametric, you can specify the number of permutations to perform to infer the distribution of the null hypothesis
Pairing
Define if the data is paired or unpaired. If the data is paired, make sure that the paired files have the same alphanumerical rank in the two folders, otherwise R will load matrices that do not correspond
Alternative
Define the H1 hypothesis (Group1 != Group2, Group1 > Group2 or Group1 < Group2)
Alpha1
Threshold to detect significantly different sub-networks
Alpha2
Threshold bellow which the null hypothesis Group1 == Group2 can be rejected in favor of H1 after correction for multiple testing
Strong
Factor by which the p-values of non-significatively different sub-networks should be multiplied in the RMIO procedure
Plot results
Check to save plots of the connections showing significantly different subnetwork measures