def get_results(randomise_outputs):
"""
Takes a list of randomise image outputs and loads them into a
dictionary containing numpy array for further processing
Parameters:
-----------
randomise_outputs : list
list of files output by randomise (ex. *vox_p_tstat1.nii.gz)
Returns:
--------
results : dict
key = contrast name (ex. 'contrast1')
value = numpy array containing 1 - p values
of group comparison output by randomise
"""
results = {}
for i in range(len(randomise_outputs)):
pth, fname, ext = gift_utils.split_filename(randomise_outputs[i])
data_1minuspval = load_rand_img(randomise_outputs[i])
data_pval = 1 - data_1minuspval
outfile = os.path.join(pth, fname + '.txt')
np.savetxt(outfile,
data_pval,
fmt='%1.5f',
delimiter='\t')
print('p value matrix saved to %s'%(outfile))
conname = ''.join(['contrast', str(i+1)])
results[conname] = data_pval
return results
# Save out text files of correlation and lags
fnc_corr_outfile = os.path.join(datadir, fnc_corr_out)
np.savetxt(fnc_corr_outfile, fnc_corr, fmt='%1.5f', delimiter='\t')
fnc_corr_z_outfile = os.path.join(datadir, fnc_corr_z_out)
np.savetxt(fnc_corr_z_outfile, fnc_corr_z, fmt='%1.5f', delimiter='\t')
fnc_lag_outfile = os.path.join(datadir, fnc_lag_out)
np.savetxt(fnc_lag_outfile, fnc_lag, fmt='%1.2f', delimiter='\t')
## Run group analysis
#######################
exists, resultsdir = gu.make_dir(datadir,'randomise')
resultsglob = os.path.join(datadir, 'FNCtb_*.csv')
result_files = glob(resultsglob)
for fnc_data_file in result_files:
fnc_data = np.genfromtxt(fnc_data_file, names=None, dtype=float, delimiter=None)
pth, fname, ext = gu.split_filename(fnc_data_file)
fnc_img_fname = os.path.join(resultsdir, fname + '.nii.gz')
fnc_saveimg = gu.save_img(fnc_data, fnc_img_fname)
rand_basename = os.path.join(resultsdir, fname)
p_uncorr_list, p_corr_list = ga.randomise(fnc_saveimg,
rand_basename,
des_file,
con_file)
uncorr_results = ga.get_results(p_uncorr_list)
corr_results = ga.get_results(p_corr_list)
fdr_results = {}
for i in range(len(uncorr_results.keys())):
conname = sorted(uncorr_results.keys())[i]
fdr_corr_arr = ga.multi_correct(uncorr_results[conname])
fdr_results[conname] = gu.square_from_combos(fdr_corr_arr, nnodes)
