#!/usr/bin/env python
import numpy as np
import cPickle
import multi_stats
import sys
if __name__ == "__main__":
if len(sys.argv) < 2:
print("Usage: show_comp_count.py <comp_count_file> [Nhyp]")
sys.exit(1)
print("Loading file %s ..." % sys.argv[1])
with open(sys.argv[1], "r") as f:
data = cPickle.load(f)
Nhyp = int(sys.argv[2]) if len(sys.argv) >= 3 else None
# load data and surrogate eigenvalues
dlam = data["dlam"]
slam = data["slam_ar"]
p_vals = multi_stats.compute_eigvals_pvalues(dlam, slam)
Nsurr = slam.shape[0]
print("Bonferroni (single-level): %d components." % np.sum(multi_stats.bonferroni_test(p_vals, 0.05, Nsurr, Nhyp)))
print("Sidak (single-level): %d components." % np.sum(multi_stats.sidak_test(p_vals, 0.05, Nsurr, Nhyp)))
print("Bonferroni-Holm (stepdown): %d components." % np.sum(multi_stats.holm_test(p_vals, 0.05, Nsurr, Nhyp)))
print("False Discovery Rate: %d components." % np.sum(multi_stats.fdr_test(p_vals, 0.05, Nsurr, Nhyp)))
'orders' : sgf.model_orders()}, f)
elif USE_MUVAR:
with open('results/hgt500_all_var_muvar_freq_comp_count_cosweights_pilot.bin', 'w') as f:
cPickle.dump({ 'dlam' : dlam, 'slam_ar' : slam_ar, 'slam_w1' : slam_w1, 'slam_f' : slam_f,
'orders' : sgf.model_orders()}, f)
else:
with open('results/hgt500_all_var_data_freq_comp_count_cosweights_pilot.bin', 'w') as f:
cPickle.dump({ 'dlam' : dlam, 'slam_ar' : slam_ar, 'slam_w1' : slam_w1, 'slam_f' : slam_f,
'orders' : sgf.model_orders()}, f)
# <codecell>
figure(figsize = (10,6))
FROM_EIG = 1
TO_EIG = 10
FR = 0
x = np.arange(TO_EIG - FROM_EIG) + FROM_EIG
plot(x, dlam[x-1,FR], 'ro-')
errorbar(x, np.mean(slam_ar[:, x-1,FR], axis = 0), np.std(slam_ar[:, x-1,FR] * 3, axis = 0), fmt = 'bo-')
errorbar(x, np.mean(slam_w1[:, x-1,FR], axis = 0), np.std(slam_w1[:, x-1,FR] * 3, axis = 0), fmt = 'go-')
errorbar(x, np.mean(slam_f[:, x-1,FR], axis = 0), np.std(slam_f[:, x-1,FR] * 3, axis = 0), fmt = 'ko-')
legend(('Data', 'AR', 'WN', 'F1'))
title('Eigenvalues for frequency range %d' % FR)
# <codecell>
print slam_ar.shape[0]
pvals = compute_eigvals_pvalues(dlam, slam_ar)
print("Bonferroni correction: %d significant components." % np.sum(bonferroni_test(pvals, 0.05, slam_ar.shape[0])))
print("FDR correction: %d significant components." % np.sum(fdr_test(pvals, 0.05, slam_ar.shape[0])))
figure(figsize=(10, 6))
FROM_EIG = 1
TO_EIG = 10
FR = 0
x = np.arange(TO_EIG - FROM_EIG) + FROM_EIG
plot(x, dlam[x - 1, FR], 'ro-')
errorbar(x,
np.mean(slam_ar[:, x - 1, FR], axis=0),
np.std(slam_ar[:, x - 1, FR] * 3, axis=0),
fmt='bo-')
errorbar(x,
np.mean(slam_w1[:, x - 1, FR], axis=0),
np.std(slam_w1[:, x - 1, FR] * 3, axis=0),
fmt='go-')
errorbar(x,
np.mean(slam_f[:, x - 1, FR], axis=0),
np.std(slam_f[:, x - 1, FR] * 3, axis=0),
fmt='ko-')
legend(('Data', 'AR', 'WN', 'F1'))
title('Eigenvalues for frequency range %d' % FR)
# <codecell>
print slam_ar.shape[0]
pvals = compute_eigvals_pvalues(dlam, slam_ar)
print("Bonferroni correction: %d significant components." %
np.sum(bonferroni_test(pvals, 0.05, slam_ar.shape[0])))
print("FDR correction: %d significant components." %
np.sum(fdr_test(pvals, 0.05, slam_ar.shape[0])))
'dlam': dlam,
'slam_ar': slam_ar,
'slam_w1': slam_w1,
'slam_f': slam_f,
'orders': sgf.model_orders()
}, f)
log("Saved results to file %s" % fname)
# show standard multiple-comparison tests assuming number of hypotheses equal to len(pvals)
pvals = compute_eigvals_pvalues(dlam, slam_ar)
log("Significance level is %g" % SIG_LEVEL)
log("Bonferroni correction: %d significant components." %
np.sum(bonferroni_test(pvals, SIG_LEVEL, NUM_SURR)))
log("Bonferroni-Holm correction: %d significant components." %
np.sum(holm_test(pvals, SIG_LEVEL, NUM_SURR)))
fdr_comps = np.sum(fdr_test(pvals, SIG_LEVEL, NUM_SURR))
log("FDR correction: %d significant components." % fdr_comps)
log("Rendering eigenvalues from data and from surrogates")
f = plt.figure(figsize=(12, 8))
FROM_EIG = 1
TO_EIG = fdr_comps + 5
x = np.arange(TO_EIG - FROM_EIG) + FROM_EIG
plt.plot(x, dlam[x - 1], 'ro-')
plt.errorbar(x,
np.mean(slam_ar[:, x - 1], axis=0),
np.std(slam_ar[:, x - 1] * 3, axis=0),
fmt='bo-')
plt.errorbar(x,
np.mean(slam_w1[:, x - 1], axis=0),
np.std(slam_w1[:, x - 1] * 3, axis=0),
'slam_f' : slam_f, 'orders' : sgf.model_orders()}, f)
log("Saved results to file %s" % fname)
else:
fname = 'results/%s_var_data_comp_count_cosweights%s.bin' % (DATA_NAME, SUFFIX)
with open(fname, 'w') as f:
cPickle.dump({ 'dlam' : dlam, 'slam_ar' : slam_ar, 'slam_w1' : slam_w1,
'slam_f' : slam_f, 'orders' : sgf.model_orders()}, f)
log("Saved results to file %s" % fname)
# show standard multiple-comparison tests assuming number of hypotheses equal to len(pvals)
pvals = compute_eigvals_pvalues(dlam, slam_ar)
log("Significance level is %g" % SIG_LEVEL)
log("Bonferroni correction: %d significant components." % np.sum(bonferroni_test(pvals, SIG_LEVEL, NUM_SURR)))
log("Bonferroni-Holm correction: %d significant components." % np.sum(holm_test(pvals, SIG_LEVEL, NUM_SURR)))
fdr_comps = np.sum(fdr_test(pvals, SIG_LEVEL, NUM_SURR))
log("FDR correction: %d significant components." % fdr_comps)
log("Rendering eigenvalues from data and from surrogates")
f = plt.figure(figsize = (12,8))
FROM_EIG = 1
TO_EIG = fdr_comps + 5
x = np.arange(TO_EIG - FROM_EIG) + FROM_EIG
plt.plot(x, dlam[x-1], 'ro-')
plt.errorbar(x, np.mean(slam_ar[:, x-1], axis = 0), np.std(slam_ar[:, x-1] * 3, axis = 0), fmt = 'bo-')
plt.errorbar(x, np.mean(slam_w1[:, x-1], axis = 0), np.std(slam_w1[:, x-1] * 3, axis = 0), fmt = 'go-')
plt.errorbar(x, np.mean(slam_f[:, x-1], axis = 0), np.std(slam_f[:, x-1] * 3, axis = 0), fmt = 'ko-')
plt.legend(('Data', 'AR', 'WN', 'F1'))
plt.title('Eigenvalues for data and surrogates')
plt.savefig('%s_eigvals_comparison%s.png' % (DATA_NAME, SUFFIX))
plt.close(f)
import numpy as np
import cPickle
import multi_stats
import sys
if __name__ == '__main__':
if len(sys.argv) < 2:
print("Usage: show_comp_count.py <comp_count_file> [Nhyp]")
sys.exit(1)
print("Loading file %s ..." % sys.argv[1])
with open(sys.argv[1], 'r') as f:
data = cPickle.load(f)
Nhyp = int(sys.argv[2]) if len(sys.argv) >= 3 else None
# load data and surrogate eigenvalues
dlam = data['dlam']
slam = data['slam_ar']
p_vals = multi_stats.compute_eigvals_pvalues(dlam, slam)
Nsurr = slam.shape[0]
print("Bonferroni (single-level): %d components." %
np.sum(multi_stats.bonferroni_test(p_vals, 0.05, Nsurr, Nhyp)))
print("Sidak (single-level): %d components." %
np.sum(multi_stats.sidak_test(p_vals, 0.05, Nsurr, Nhyp)))
print("Bonferroni-Holm (stepdown): %d components." %
np.sum(multi_stats.holm_test(p_vals, 0.05, Nsurr, Nhyp)))
print("False Discovery Rate: %d components." %
np.sum(multi_stats.fdr_test(p_vals, 0.05, Nsurr, Nhyp)))