def my_t_test(a, b, a_label='a', b_label='b'):
"""Given two array-like objects, calculates the P-Value"""
p_value = t_test(a, b)[1]
mean_a = round(a.mean(), 2)
mean_b = round(b.mean(), 2)
p_value = round(p_value, 4)
if p_value < 0.05:
print('The difference is statistically significant. p =', p_value)
print('Mean of array "%s": %s. N=%s' % (a_label, mean_a, len(a)))
print('Mean of array "%s": %s. N=%s' % (b_label, mean_b, len(b)))
else:
print('Cannot reject the null hypothesis. p =', p_value)
print('Mean of array "%s": %s. N=%s' % (a_label, mean_a, len(a)))
print('Mean of array "%s": %s. N=%s' % (b_label, mean_b, len(b)))
def streak_test(rolls: list, die_size: int = None):
if die_size is None:
die_size = max(rolls)
exp_diff, exp_list = expected_diff(die_size)
diffs = []
abs_diffs = []
for i in range(1, len(rolls)):
seq_diff = rolls[i] - rolls[i - 1]
diffs.append(seq_diff)
abs_diffs.append(abs(seq_diff))
mean_diff = sum(diffs) / len(diffs)
mean_abs_diff = sum(abs_diffs) / len(abs_diffs)
t_test_result = t_test(abs_diffs, exp_list)
p_val = t_test_result.pvalue
return p_val, mean_abs_diff, exp_diff
def Plot_electrode_welch(attention, motor, freqs, electrode, con1, con2, comparison_type, PLOT_ALL = True):
plt.close('all')
mean_att = np.mean(attention, axis = 0)[3:36]
mean_mot = np.mean(motor, axis = 0)[3:36]
p_vals = []
for frequency in range(3,36):
t,p = t_test(attention[:, frequency], motor[:, frequency])
p_vals.append(p)
sigs = [idx for idx,p_val in enumerate(p_vals) if p_val < 0.005]
if(PLOT_ALL):
fig = plt.figure()
fig.suptitle(ch_names[electrode])
gs = gridspec.GridSpec(2, 1, height_ratios=[2,1])
att_mot = plt.subplot(gs[0])
test = plt.subplot(gs[1])
att_mot.plot(FREQS[3:36], np.log10(mean_att) , color = 'green', label = con1)
att_mot.plot(FREQS[3:36], np.log10(mean_mot), color = 'red', label = con2)
test.plot(FREQS[3:36], p_vals, color = 'black', linestyle = '--', label = 'p value')
test.hlines(y = 0.005, xmin =FREQS[3], xmax = FREQS[36], color = 'red', linestyle = '--', label = 'alpha 0.005')
test.set_xlim(FREQS[3], FREQS[36])
att_mot.set_xlim(FREQS[3], FREQS[36])
test.set_ylabel('Independent t-test')
att_mot.set_ylabel('Welch')
test.set_xlabel('Frequency')
test.legend(loc = 'best')
att_mot.legend(loc = 'best')
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/All/' + con1 + '_' + con2
if not os.path.exists(directory):
os.makedirs(directory)
savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/All/' + con1 + '_' + con2 +'/'+ ch_names[electrode] + '.png')
if(sigs != []):
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/Sig/' + con1 + '_' + con2
if not os.path.exists(directory):
os.makedirs(directory)
savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/Sig/' + con1 + '_' + con2 +'/'+ ch_names[electrode] + '.png')
test.set_axis_bgcolor('red')
print('FOUND DIFFRENCE AT: ')
print(np.array(sigs) *2)
if((sigs != []) & (PLOT_ALL == False)):
fig = plt.figure()
fig.suptitle(ch_names[electrode])
gs = gridspec.GridSpec(2, 1, height_ratios=[2,1])
att_mot = plt.subplot(gs[0])
test = plt.subplot(gs[1])
att_mot.plot(FREQS[3:36], np.log10(mean_att) , color = 'green', label =con1)
att_mot.plot(FREQS[3:36], np.log10(mean_mot), color = 'red', label = con2)
test.plot(FREQS[3:36], p_vals, color = 'black', linestyle = '--', label = 'p value')
test.hlines(y = 0.005, xmin =FREQS[3], xmax = FREQS[36], color = 'red', linestyle = '--', label = 'alpha 0.005')
test.set_xlim(FREQS[3], FREQS[36])
att_mot.set_xlim(FREQS[3], FREQS[36])
test.set_ylabel('Independent t-test')
att_mot.set_ylabel('Welch')
test.set_xlabel('Frequency')
test.legend(loc = 'best')
att_mot.legend(loc = 'best')
test.set_axis_bgcolor('red')
print('FOUND DIFFRENCE AT: ')
print(np.array(sigs) *2)
def get_significant_groupings(groupdata, labels=None, cutoff=0.05):
print("\t\tDetermining significant letter groupings for plot\n")
if labels == None:
labels = ["Data" + str(i + 1)
for i in range(len(groupdata))] # Labels for each dataset
# Calculate p-values
pvals = np.empty(shape=(len(groupdata), len(groupdata)))
pvals[:] = np.nan
for i in range(len(groupdata)):
myset = set()
for j in range(i + 1, len(groupdata)):
t, p = t_test(groupdata[i], groupdata[j])
pvals[i, j], pvals[j, i] = p, p
pvals = pd.DataFrame(pvals, columns=labels, index=labels)
print("P-values:\n", pvals)
# Make all possible combinations of groups
allgroups = list()
for size in range(2, len(labels) + 1):
for mygroup in itertools.combinations(labels, size):
allgroups.append(mygroup)
#print(mygroup)
# Subset to just the ones where all members are statistically indistinguishable
goodgroups = list()
for mygroup in allgroups:
isgood = True
for i in range(len(mygroup)):
for j in range(i, len(mygroup)):
if pvals.loc[mygroup[i], mygroup[j]] <= cutoff:
isgood = False
#print(isgood, mygroup)
if isgood: goodgroups.append(mygroup)
# Check if one-member groups exist by collapsing groups to a set of labels
already_included = set(np.hstack(goodgroups))
for l in labels:
if l not in already_included:
goodgroups.append([l]) #Append 1-item list of groups
#print(goodgroups)
# Now assign letter values; do in a loop over labels to preserve order
lettercodes = {l: ""
for l in labels
} # Dictionary of which letter codes each label has
letter_i = 0 # Keeping track of which group label we're using
for l in labels:
for g in goodgroups:
if l not in g: continue # Skip groups this label is not in
for mylabel in g:
lettercodes[mylabel] += string.ascii_lowercase[letter_i]
goodgroups.remove(
g
) # Remove a group after it's been processed so it doesn't get processed a second time
letter_i += 1
#print(lettercodes)
# Make letters in order of original list
lettervals = [lettercodes[l] for l in labels]
#print(lettervals)
return lettervals, pvals
def Plot_electrode_welch(all_results, training, event_type, conditions):
"""Takes for an input a dictionary which has two fields, electrodes_a and b. event_type is a string. condition is a list of strings describing what comparison is made"""
#plt.close('all')
for (name_a, electrode_a), (name_b, electrode_b) in zip(all_results[conditions[0]].items(), all_results[conditions[1]].items()):
fig = plt.figure()
fig.suptitle(name_a + '\n'+ conditions[0] + ' vs ' + conditions[1] + '\n' + event_type, fontweight = 'bold')
a_b_conditions = fig.add_subplot(111)
#Iterate in two seperate loops since they might (and in fact are for sure) of unequal lengths
_min = np.argmax(FREQS > 0)
_max = np.argmax(FREQS > 30)
for trial_a in electrode_a:
#just reshaping
normed_a = trial_a[_min:_max] / trial_a[_min:_max].mean()
a_b_conditions.plot(FREQS[_min: _max], np.log10(normed_a), color = 'green', alpha = 0.1)
for trial_b in electrode_b:
#just reshaping
normed_b = trial_b[_min:_max] / trial_b[_min:_max].mean()
a_b_conditions.plot(FREQS[_min:_max], np.log10(normed_b), color = 'red', alpha = 0.1)
# a_b_conditions.set_xlim(min(FREQS), max(FREQS))
a_b_conditions.set_xlabel('Frequency')
a_b_conditions.set_ylabel('Welch power spectrum')
try:
mean_a = electrode_a.mean(axis = 0)#[_min:_max]
mean_b = electrode_b.mean(axis = 0)
# print(mean_a)
a_b_conditions.plot(FREQS[_min:_max], np.log10(mean_a[_min:_max]/mean_a[_min:_max].mean()) ,label = conditions[0], linewidth=2, color = 'green')
a_b_conditions.plot(FREQS[_min:_max], np.log10(mean_b[_min:_max]/mean_b[_min:_max].mean()) ,label = conditions[1], linewidth=2, color = 'red')
p_vals = []
for frequency in range(0, len(FREQS)):
t,p = t_test(electrode_a[:, frequency], electrode_b[:, frequency])
p_vals.append(p)
sigs = [idx for idx,p_val in enumerate(p_vals) if p_val < 0.005]
#save in all
for sig in sigs:
#sig times two because the index is half the actuall frequency
a_b_conditions.vlines(x = sig*2, ymin = -2.5, ymax = 1.5, color = 'blue', linestyle = '--', alpha = 0.2)
# test.hlines(y = 0.005, xmin =FREQS[3], xmax = FREQS[36], color = 'red', linestyle = '--', label = 'alpha 0.005')
a_b_conditions.legend(loc = 'best')
path_conditions = conditions[0] + '_' + conditions[1]
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/Independent/'+ event_type +'/' + path_conditions + '/'+ training + '/All/'
if not os.path.exists(directory):
os.makedirs(directory)
plt.savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/Independent/'+ event_type +'/' + path_conditions+ '/'+ training + '/All/'+ name_a + '.png')
#Save also in sigs
if sigs != []:
directory_sig = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/Independent/'+ event_type +'/' + path_conditions+ '/'+ training + '/Sig/'
if not os.path.exists(directory_sig):
os.makedirs(directory_sig)
plt.savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/Independent/'+ event_type +'/' + path_conditions + '/'+ training + '/Sig/'+ name_a + '.png')
except:
# print(a_avg)
#print(b_avg)
print('Probably not enouth trials')
def Plot_electrode_welch(_all_results, event_type, condition):
plt.close('all')
#event type, i.e. databse, CUe or Target, conditon, for example 'attention correct
#This loop is to reverse the nesting from subject * electrodes to the other way
all_electrodes_before_and_after = {name: [] for name in ch_names}
for subject in _all_results:
for before_electrode, after_electrode, electrode_name in zip(subject['electrodes_before'], subject['electrodes_after'], ch_names):
all_electrodes_before_and_after[electrode_name].append({'before':before_electrode, 'after':after_electrode})
for key, electrode in all_electrodes_before_and_after.items():
fig = plt.figure()
fig.suptitle(key, fontweight = 'bold')
before_after = plt.subplot()
before_avg, after_avg = [],[]
for idx, subject_avg in enumerate(electrode):
before_after.plot(FREQS, np.log10(subject_avg['after']), color = 'red', label = 'after', alpha = 0.1)
before_after.plot(FREQS, np.log10(subject_avg['before']), color = 'green', label = 'before', alpha = 0.1)
before_avg.append(subject_avg['before'])
after_avg.append(subject_avg['after'])
before_avg = np.array(before_avg)
after_avg = np.array(after_avg)
before_after.plot(FREQS, np.log10(before_avg.mean(axis = 0)), linewidth=2, color = 'green')
before_after.plot(FREQS, np.log10(after_avg.mean(axis = 0)), linewidth=2, color = 'red')
before_after.set_xlim(min(FREQS), max(FREQS))
before_after.set_xlabel('Frequency')
before_after.set_ylabel('Welch power spectrum')
p_vals = []
for frequency in range(0, len(FREQS)):
t,p = t_test(before_avg[:, frequency], after_avg[:, frequency])
p_vals.append(p)
sigs = [idx for idx,p_val in enumerate(p_vals) if p_val < 0.005]
#save in all
for sig in sigs:
#sig times two because the index is half the actuall frequency
before_after.vlines(x = sig*2, ymin = -2.5, ymax = 1.5, color = 'blue', linestyle = '--', label = 'alpha 0.005', alpha = 0.2)
# test.hlines(y = 0.005, xmin =FREQS[3], xmax = FREQS[36], color = 'red', linestyle = '--', label = 'alpha 0.005')
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/'+ event_type +'/' + condition + '/All/'
if not os.path.exists(directory):
os.makedirs(directory)
plt.savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/'+ event_type +'/' + condition + '/All/'+ key + '.png')
#Save also in sigs
if sigs != []:
directory_sig = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/'+ event_type +'/' + condition + '/Sig/'
if not os.path.exists(directory_sig):
os.makedirs(directory_sig)
plt.savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/'+ event_type +'/' + condition + '/Sig/'+ key + '.png')
def Plot_electrode_welch(_all_results, event_type, condition):
plt.close('all')
#event type, i.e. databse, CUe or Target, conditon, for example 'attention correct
#This loop is to reverse the nesting from subject * electrodes to the other way
all_electrodes_before_and_after = {name: [] for name in ch_names}
for subject in _all_results:
for before_electrode, after_electrode, electrode_name in zip(
subject['electrodes_before'], subject['electrodes_after'],
ch_names):
all_electrodes_before_and_after[electrode_name].append({
'before':
before_electrode,
'after':
after_electrode
})
for key, electrode in all_electrodes_before_and_after.items():
fig = plt.figure()
fig.suptitle(key, fontweight='bold')
before_after = plt.subplot()
before_avg, after_avg = [], []
for idx, subject_avg in enumerate(electrode):
before_after.plot(FREQS,
np.log10(subject_avg['after']),
color='red',
label='after',
alpha=0.1)
before_after.plot(FREQS,
np.log10(subject_avg['before']),
color='green',
label='before',
alpha=0.1)
before_avg.append(subject_avg['before'])
after_avg.append(subject_avg['after'])
before_avg = np.array(before_avg)
after_avg = np.array(after_avg)
before_after.plot(FREQS,
np.log10(before_avg.mean(axis=0)),
linewidth=2,
color='green')
before_after.plot(FREQS,
np.log10(after_avg.mean(axis=0)),
linewidth=2,
color='red')
before_after.set_xlim(min(FREQS), max(FREQS))
before_after.set_xlabel('Frequency')
before_after.set_ylabel('Welch power spectrum')
p_vals = []
for frequency in range(0, len(FREQS)):
t, p = t_test(before_avg[:, frequency], after_avg[:, frequency])
p_vals.append(p)
sigs = [idx for idx, p_val in enumerate(p_vals) if p_val < 0.005]
#save in all
for sig in sigs:
#sig times two because the index is half the actuall frequency
before_after.vlines(x=sig * 2,
ymin=-2.5,
ymax=1.5,
color='blue',
linestyle='--',
label='alpha 0.005',
alpha=0.2)
# test.hlines(y = 0.005, xmin =FREQS[3], xmax = FREQS[36], color = 'red', linestyle = '--', label = 'alpha 0.005')
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/' + event_type + '/' + condition + '/All/'
if not os.path.exists(directory):
os.makedirs(directory)
plt.savefig(
'/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/' +
event_type + '/' + condition + '/All/' + key + '.png')
#Save also in sigs
if sigs != []:
directory_sig = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/' + event_type + '/' + condition + '/Sig/'
if not os.path.exists(directory_sig):
os.makedirs(directory_sig)
plt.savefig(
'/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/BeforeAfter/' +
event_type + '/' + condition + '/Sig/' + key + '.png')
def Plot_electrode_welch(attention,
motor,
freqs,
electrode,
con1,
con2,
comparison_type,
PLOT_ALL=True):
plt.close('all')
mean_att = np.mean(attention, axis=0)[3:36]
mean_mot = np.mean(motor, axis=0)[3:36]
p_vals = []
for frequency in range(3, 36):
t, p = t_test(attention[:, frequency], motor[:, frequency])
p_vals.append(p)
sigs = [idx for idx, p_val in enumerate(p_vals) if p_val < 0.005]
if (PLOT_ALL):
fig = plt.figure()
fig.suptitle(ch_names[electrode])
gs = gridspec.GridSpec(2, 1, height_ratios=[2, 1])
att_mot = plt.subplot(gs[0])
test = plt.subplot(gs[1])
att_mot.plot(FREQS[3:36],
np.log10(mean_att),
color='green',
label=con1)
att_mot.plot(FREQS[3:36], np.log10(mean_mot), color='red', label=con2)
test.plot(FREQS[3:36],
p_vals,
color='black',
linestyle='--',
label='p value')
test.hlines(y=0.005,
xmin=FREQS[3],
xmax=FREQS[36],
color='red',
linestyle='--',
label='alpha 0.005')
test.set_xlim(FREQS[3], FREQS[36])
att_mot.set_xlim(FREQS[3], FREQS[36])
test.set_ylabel('Independent t-test')
att_mot.set_ylabel('Welch')
test.set_xlabel('Frequency')
test.legend(loc='best')
att_mot.legend(loc='best')
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/All/' + con1 + '_' + con2
if not os.path.exists(directory):
os.makedirs(directory)
savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' +
comparison_type + '/All/' + con1 + '_' + con2 + '/' +
ch_names[electrode] + '.png')
if (sigs != []):
directory = '/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' + comparison_type + '/Sig/' + con1 + '_' + con2
if not os.path.exists(directory):
os.makedirs(directory)
savefig('/Users/ryszardcetnarski/Desktop/Nencki/TD/Figs/' +
comparison_type + '/Sig/' + con1 + '_' + con2 + '/' +
ch_names[electrode] + '.png')
test.set_axis_bgcolor('red')
print('FOUND DIFFRENCE AT: ')
print(np.array(sigs) * 2)
if ((sigs != []) & (PLOT_ALL == False)):
fig = plt.figure()
fig.suptitle(ch_names[electrode])
gs = gridspec.GridSpec(2, 1, height_ratios=[2, 1])
att_mot = plt.subplot(gs[0])
test = plt.subplot(gs[1])
att_mot.plot(FREQS[3:36],
np.log10(mean_att),
color='green',
label=con1)
att_mot.plot(FREQS[3:36], np.log10(mean_mot), color='red', label=con2)
test.plot(FREQS[3:36],
p_vals,
color='black',
linestyle='--',
label='p value')
test.hlines(y=0.005,
xmin=FREQS[3],
xmax=FREQS[36],
color='red',
linestyle='--',
label='alpha 0.005')
test.set_xlim(FREQS[3], FREQS[36])
att_mot.set_xlim(FREQS[3], FREQS[36])
test.set_ylabel('Independent t-test')
att_mot.set_ylabel('Welch')
test.set_xlabel('Frequency')
test.legend(loc='best')
att_mot.legend(loc='best')
test.set_axis_bgcolor('red')
print('FOUND DIFFRENCE AT: ')
print(np.array(sigs) * 2)
