# should i zscore? lidx = np.arange(32) pidx = np.arange(32, 64) lres = cv(betas[lidx].copy()) lresults.append(lres) print "language: " + str(np.mean(lres.samples)) cv.untrain() pres = cv(betas[pidx].copy()) presults.append(pres) print "pictures: " + str(np.mean(pres.samples)) cv.untrain() fclf.train(betas[lidx].copy()) l2presults.append(np.mean(fclf.predict(betas[pidx]) == betas[pidx].sa.targets)) fclf.untrain() fclf.train(betas[pidx]) p2lresults.append(np.mean(fclf.predict(betas[lidx]) == betas[lidx].sa.targets)) import matplotlib.pyplot as plt def plotsubs(lr, pr, l2p, p2l, c=None, title=None, bar_width=.2, opacity=.4, error_config={'ecolor': '0.3'}): # results is the concatenated output of cv across subjects... or something. f, (ax1, ax2) = plt.subplots(2, figsize=(12,6)) index = np.arange(len(lr)) lheights = [] lerrbars = [] pheights = [] perrbars = [] for i in lr: