Running CWAS with Python

demo.py

# To use, please install CPAC (CWAS branch) or clone and add this to your path
#   git clone https://github.com/FCP-INDI/C-PAC.git
#   cd C-PAC
#   git checkout cwas
# also see: https://github.com/FCP-INDI/C-PAC/tree/cwas

from CPAC.cwas import calc_cwas

# Load your data
## Create the (`S`,`T`,`V`) dataset to hold everything: `S` subjects, `T` timepoints, `V` voxels
subjects_data = np.zeros((nsubs,ntpts,nvoxs))
## Loop through each subjects data, load and add to dataset.
## Data should be space separated files, although can be anything if set loadtxt differently
for i,fname in enumerate(subject_fnames):
  subjects_data[i,:,:] = np.loadtxt(fname)

# Load regressors/model
## also should be space separated file but can use something else (e.g., a csv with the pandas library)
## Matrix of shape (`S`, `R`), `S` subjects and `R` regressors
regressors = np.loadtxt(model_fname)

# Setup other variables
## indicate the columns that are your regressors of interest and you want to permute
## for instance the 2nd column might be age effects, in which case you specify 1
## (since python is 0-based, the 1 is actually the 2nd column)
cols = [1] # must be a list
## memory limit (beta)
memlimit = 4
## strata (if you have a within-subjects design, then specify this)
## for instance, if you have drug and placebo scans, then this should be a list indicating the subjects ids for each scan
## in this way, MDMR will only permute scan type (drug or placebo) within each subject
strata = None
## number of permutations
nperms = 1000

# Run
## the F_set are the raw F-statistics
## the p_set output is likely what you would be interested in giving the significance
## the Fperms are the F-stats for all the permuted datasets as well as the original data
F_set, p_set, Fperms = calc_cwas(subjects_data, regressors, cols, nperms, memlimit=memlimit, strata=strata):