Christophe@pallier.org

Before anything, set the ROOT_DIR of the project, for example:

. setroot-neurospin

To perform an analysis, first select a model in models to set several environment variables:

. setmodel models/en/chrmodels/en/rms-wordrate-freq-bottomup/

Then, you can execute the analysis step by step, following the stages in the Makefile:

make regressors        # check outputs/regressors
make design-matrices   # check outputs/design-matrices
make first-level       # check output/results-indiv
make second-level      # check output/results-group
make roi-analyses

To create a new model, you need to:

1. create a subdirectory inside the models directory

2. create a setenv file exporting the environment variables specifing the model name, the list of regressors, etc. (see modesl/en/christophe-bottomup/setenv fro an example)

3. In your model's directory, create firstlevel.py and optionaly, orthonormalize.py which will be executed by make first-level. Create also group.py for the second level.

4. If your model includes variables that have not yet been used in previous models, you need to add to the folder inputs/onsets one comma separated (.csv) file per variable and per run --- the filename pattern being X_VARNAME.csv where X is the run number [1-9]. Each file must contain two columns named 'onset' and 'amplitude' (onsets is given in seconds).

Requirements:

• Python: pandas, nistats, nibabel, nilearn, statsmodels