Permutation testing for non-imaging data using FSL randomise

The randomise_non_imaging script is designed to take advantage of the functionalities of FSL randomise to perform GLM-based non-parametric permutation testing using non-imaging data. This can be done fairly easily with other programs, but using randomise could be convenient to FSL users, who are accustomed to creating the necessary input files.

How to make it work
Download from here.
System requirements:
This scripts requires FSL as well as python (including the numpy and nipy packages), and is meant to be used in Linux systems. The necessary python packages can be easily obtained by installing Anaconda.
Add alias and route to .bashrc:
After unzipping, we recommend adding an alias to the user’s .bashrc as an easy way to call the script from any terminal. The path to the folder containing the parameter2nifti.pyscript should also be specified as route_NIR in the .bashrc:
alias randomise_non_imaging='bash /full/path/to/your/folder/'

export route_NIR=/full/path/to/your/folder/
Input files:
Three basic input files are required:
1. Dependent variable matrix: text file containing the variables to be tested, consisting of one column per variable and one row for each observation. This is equivalent to the image input in randomise – and will in fact be converted to image format so it can be fed into the program
2. Design matrix (.mat)
3. Contrast matrix (.con): no more than 9 contrasts should be included in each contrast matrix.
Some options require additional input files:
1. F tests: requires .fts files (with the same root name as the design and contrast files)
2. Block permutation: requires exchangeability block labels .grp file (with the same root name as the design and contrast files)
Output (text) files:
1. P value file: named (output)_p_all_contrasts
2. Stats file: named (output)_stat_all_contrasts
3. F test p value file: (output)_p_F_test
4. F test stats file: (output)_Fstat
5. Corrected p value file: (output)_corrp_all_contrasts
6. Corrected F test p value file: (output)_corrp_F_test
* We recommend creating the .mat, .con, .fts, and .grp files using the Glm tool in FSL. Details can be found here.
randomise_non_imaging -V (V_input) -D (D_input) -O (O_name) [options]
Compulsory arguments:
-V: dependent variable matrix
-D: root name of the design (.mat), contrast (.con), and (optionally) F-test (.fts) and exchangeability block (.grp) files
-O: output root name
Optional arguments:
– F: perform F-test. Options are YES or NO (default is NO). If set to YES, the .fts file must be present
-G: perform block permutation. Options are YES or NO (default is NO). If set to YES, the .grp file must be present
-C: By default, p values are not corrected for multiple testing. Optionally, family-wise error-corrected p values can be generated (option -C YES) by comparing the statistic of each variable to the maximum statistic distribution across all variables under the null hypothesis.
-P: The number of permutations performed (default 10,000) can be specified using the -P option.
Example data set:
Download and unzip This file contains a variable file (MMSE_BDI), corresponding to two test scores for 72 subjects. The design matrix control_vs_patient.mat specifies two subject groups (controls and patients), and the contrast matrix control_vs_patient.con specifies two contrasts (control>patient, patient>control). The corresponding randomise_non_imaging command would be:
randomise_non_imaging -V MMSE_BDI -D control_vs_patient -O intergroup_comparison
Hope this can be of use!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s