def test_comparison_not_equal_after_iteration(self):
q1 = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=1)
q2 = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=2)
q1.__next__()
q2.__next__()
assert q1 != q2
def run_quest_procedure(self,
n_trials=10,
start=20,
start_sd=10,
maxvalue=40,
beta=3.5,
delta=0.01,
gamma=0.01,
grain=0.01):
print("RUNNING QUEST\n" \
"-------------\n")
self.quest = data.QuestHandler(start,
start_sd,
pThreshold=0.75,
nTrials=n_trials,
minVal=0,
maxVal=maxvalue,
beta=beta,
delta=delta,
gamma=gamma,
grain=grain)
for n, stim_level in enumerate(self.quest):
print("Calibration trial {0} / {1}\n" \
"Stimulation level = {2}\n" \
"-----------------------------".format(n + 1, n_trials, stim_level))
response = None
# Repeat if subject doesn't respond
while response == None:
trial = ConditioningTrial(self,
n,
True,
max_voltage=stim_level)
response = trial.run()
# Add the response
if response:
print("Stimulation detected")
else:
print("Stimulation not detected")
if n >= self.config['quest_settings']['n_ignored']:
self.quest.addResponse(int(response))
p0 = self.quest.quantile(0.01)
p25 = self.quest.quantile(.25)
p50 = self.quest.quantile(.50)
p75 = self.quest.quantile(.75)
print("1% detection probability level = {0}\n"
"25% detection probability level = {1}\n"
"50% detection probability level = {2}\n"
"75% detection probability level = {3}\n".format(
self.quest.quantile(0.01), self.quest.quantile(.25),
self.quest.quantile(.5), self.quest.quantile(.75)))
return p0, p25, p50, p75
def test_json_dump(self):
q = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=1)
dump = q.saveAsJson()
q.origin = ''
assert q == json_tricks.np.loads(dump)
def test_comparison_not_equal(self):
q1 = data.QuestHandler(0.5,
0.2,
pThreshold=0.63,
gamma=0.01,
nTrials=20,
minVal=0,
maxVal=1)
q2 = data.QuestHandler(0.5,
0.2,
pThreshold=0.63,
gamma=0.01,
nTrials=20,
minVal=0,
maxVal=2)
assert q1 != q2
def test_epsilon(self):
# Values used by Harvey (1986), Table 3.
beta = 3.5
gamma = 0
delta = 0
epsilon = 0.0571
# QuestHandler needs this, but it doesn't influence our
# test. Values chosen arbitrarily.
startVal = 0.5
startValSd = 1
# Estimate the target proportion correct based on epsilon.
# (The values provided by Harvey (1986) are rounded to two
# decimal places and, therefore, too imprecise. So we
# have to we calculate it again.)
def weibull(x, beta, gamma, delta):
p = delta * gamma + (1 - delta) * (
1 - (1 - gamma) * np.exp(-10**(beta * x)))
return p
p = weibull(x=epsilon, beta=beta, gamma=gamma, delta=delta)
q = data.QuestHandler(startVal=startVal,
startValSd=startValSd,
pThreshold=p,
beta=beta,
gamma=gamma,
delta=delta)
assert np.isclose(q.epsilon, epsilon, atol=1e-4)
def test_QuestHandler(self):
nTrials = 10
startVal, minVal, maxVal = 50, 0, 100
range = maxVal - minVal
startValSd = 50
grain = 0.01
pThreshold = 0.82
beta, gamma, delta = 3.5, 0.5, 0.01
stopInterval = None
method = 'quantile'
self.stairs = data.QuestHandler(startVal,
startValSd,
pThreshold=pThreshold,
nTrials=nTrials,
stopInterval=stopInterval,
method=method,
beta=beta,
gamma=gamma,
delta=delta,
grain=grain,
range=range,
minVal=minVal,
maxVal=maxVal)
self.stairs.nReversals = None
self.responses = makeBasicResponseCycles(cycles=3,
nCorrect=2,
nIncorrect=2,
length=10)
self.intensities = [
50, 45.139710407074872, 37.291086503930742, 58.297413127139947,
80.182967131096547, 75.295251409003527, 71.57627192423783,
79.881680484036906, 90.712313302815517, 88.265808957695796
]
mean = 86.0772169427
mode = 80.11
quantile = 86.3849031085
self.simulate()
self.checkSimulationResults()
assert self.stairs.startVal == startVal
assert self.stairs.startValSd == startValSd
assert np.allclose(self.stairs.mean(), mean)
assert np.allclose(self.stairs.mode(), mode)
assert np.allclose(self.stairs.quantile(), quantile)
# Check if the internal grid has the expected dimensions.
assert len(self.stairs._quest.x) == (maxVal - minVal) / grain + 1
assert np.allclose(self.stairs._quest.x[1] - self.stairs._quest.x[0],
grain)
assert self.stairs._quest.x[0] == -range / 2
assert self.stairs._quest.x[-1] == range / 2
def test_json_dump_with_data(self):
q = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=1)
q.addResponse(1)
q.addOtherData('foo', 'bar')
dump = q.saveAsJson()
q.origin = ''
assert q == json_tricks.np.loads(dump)
def test_json_dump_to_file(self):
_, path = mkstemp(dir=self.tmp_dir, suffix='.json')
q = data.QuestHandler(0.5,
0.2,
pThreshold=0.63,
gamma=0.01,
nTrials=20,
minVal=0,
maxVal=1)
q.saveAsJson(fileName=path, fileCollisionMethod='overwrite')
def get_blocks(self):
blocks = []
for name in self.oris.keys():
blocks.append(
data.QuestHandler(15,
3,
pThreshold=.63,
nTrials=self.ntrials_diff,
minVal=0,
maxVal=25,
name=name))
self.blocks = blocks
def test_json_dump_and_reopen_file(self):
q = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=1)
q.addResponse(1)
q.addOtherData('foo', 'bar')
q.__next__()
_, path = mkstemp(dir=self.tmp_dir, suffix='.json')
q.saveAsJson(fileName=path, fileCollisionMethod='overwrite')
q.origin = ''
q_loaded = fromFile(path)
assert q == q_loaded
def test_attributes(self):
beta = 3.5
gamma = 0.01
delta = 0.02
grain = 0.01
range = 1
q = data.QuestHandler(startVal=0.5,
startValSd=0.2,
pThreshold=0.63,
nTrials=10,
beta=beta,
gamma=gamma,
delta=delta,
grain=grain,
range=range)
assert q.beta == beta
assert q.gamma == gamma
assert q.delta == delta
assert q.grain == grain
assert q.range == range
def run_session(self):
path = os.path.join(self.res_dir, self.subject)
if not os.path.exists(path):
os.makedirs(path)
# welcome
msg = visual.TextStim(self.win,
'Welcome!' + '\n' +
' Press any key to start this session :)',
color='black',
units='deg',
pos=(0, 0),
height=0.8)
msg.draw()
self.win.mouseVisible = False
self.win.flip()
event.waitKeys()
# read staircase parameters
conditions = [
dict({'stimulus': key}, **value)
for key, value in self.param.items() if key.startswith('stimulus')
]
if conditions[0]['stairType'] == 'simple':
stairs = data.MultiStairHandler(stairType='simple',
conditions=conditions,
nTrials=self.trial_nmb,
method='sequential')
elif conditions[0]['stairType'] == 'quest':
stairs = []
for cond in conditions:
if self.priors_file_path:
prior_file = self.priors_file_path + cond[
'label'] + '.psydat'
print(prior_file)
prior_handler = misc.fromFile(prior_file)
else:
prior_handler = None
cur_handler = data.QuestHandler(cond['startVal'],
cond['startValSd'],
pThreshold=cond['pThreshold'],
nTrials=self.trial_nmb,
minVal=cond['min_val'],
maxVal=cond['max_val'],
staircase=prior_handler,
extraInfo=cond)
stairs.append(cur_handler)
elif conditions[0]['stairType'] == 'psi':
stairs = []
for cond in conditions:
if self.priors_file_path:
prior_file = self.priors_file_path + cond['label'] + '.npy'
else:
prior_file = None
print(prior_file)
cur_handler = data.PsiHandler(nTrials=self.trial_nmb,
intensRange=[1, 10],
alphaRange=[1, 10],
betaRange=[0.01, 10],
intensPrecision=0.1,
alphaPrecision=0.1,
betaPrecision=0.01,
delta=0.01,
extraInfo=cond,
prior=prior_file,
fromFile=(prior_file
is not None))
stairs.append(cur_handler)
# write configuration files
xpp = config_tools.WriteXpp(self.subject, self.idx)
xpp_file = xpp.head(self.cfg_file, self.par_file)
config_tools.write_xrl(self.subject,
cfg_file=self.cfg_file,
par_file=self.par_file,
xpp_file=xpp_file)
xlsname = path + '/' + self.idx + self.param[
'noise_condition'] + '.xlsx'
""" running staircase """
if isinstance(stairs, data.MultiStairHandler):
# start running the staircase using the MultiStairHandler for the up-down method
count = 0
for rot, cond in stairs:
count += 1
judge, thiskey, trial_time = self.run_trial(rot, cond, count)
# check whether the theta is valid - if not, the rotation given by staircase should be corrected by
# realizable values
valid_theta = np.round(np.load(self.hue_list), decimals=1)
disp_standard = self.take_closest(
valid_theta, cond['standard']) # theta actually displayed
stair_test = cond['standard'] + stairs._nextIntensity * (-1)**(
cond['label'].endswith('m')) # theta for staircase
if stair_test < 0:
stair_test += 360
disp_test = self.take_closest(valid_theta, stair_test)
disp_intensity = abs(disp_test - disp_standard)
if disp_intensity > 300:
disp_intensity = 360 - (disp_test + disp_standard)
stairs.addResponse(judge, disp_intensity)
xpp.task(count, cond, rot, float(disp_intensity), judge,
trial_time)
event.waitKeys(keyList=[
thiskey
]) # press the response key again to start the next trial
config_tools.write_xrl(self.subject, xls_file=xlsname)
stairs.saveAsExcel(xlsname) # save results
psydat_file_path = os.path.join(
path, "psydat", self.idx + self.param['condition'] +
'.psydat') # save the handler into a psydat-file
misc.toFile(psydat_file_path, stairs)
elif isinstance(stairs, list):
# start running the staircase using custom interleaving stairs for the quest and psi methods
count = 0
rot_all = []
rot_all_disp = []
judge_all = []
estimates = {s.extraInfo['label']: [] for s in stairs}
for trial_n in range(self.trial_nmb):
for handler_idx, cur_handler in enumerate(stairs):
count += 1
rot = next(cur_handler)
if len(rot_all) <= handler_idx:
rot_all.append([])
rot_all[handler_idx].append(rot)
cond = cur_handler.extraInfo
judge, thiskey, trial_time = self.run_trial(
rot, cond, count)
if len(judge_all) <= handler_idx:
judge_all.append([])
judge_all[handler_idx].append(judge)
# cur_handler.addResponse(judge) # to the next trial
valid_theta = np.round(np.load(self.hue_list), decimals=1)
disp_standard = self.take_closest(
valid_theta,
cond['standard']) # theta actually displayed
stair_test = cond[
'standard'] + cur_handler._nextIntensity * (-1)**(
cond['label'].endswith('m')) # theta for staircase
if stair_test < 0:
stair_test += 360
disp_test = self.take_closest(valid_theta, stair_test)
disp_intensity = abs(disp_test - disp_standard)
if disp_intensity > 300:
disp_intensity = 360 - (disp_test + disp_standard)
cur_handler.addResponse(judge, disp_intensity)
if len(rot_all_disp
) <= handler_idx: # add displayed intensities
rot_all_disp.append([])
rot_all_disp[handler_idx].append(disp_intensity)
print('stair test: ' + str(stair_test) + ', ' +
'disp_test:' + str(disp_test))
if isinstance(cur_handler, data.PsiHandler):
estimates[cur_handler.extraInfo['label']].append([
cur_handler.estimateLambda()[0], # location
cur_handler.estimateLambda768()[1], # slope
cur_handler.estimateThreshold(0.75)
])
elif isinstance(cur_handler, data.QuestHandler):
estimates[cur_handler.extraInfo['label']].append([
cur_handler.mean(),
cur_handler.mode(),
cur_handler.quantile(0.5)
])
xpp.task(count, cond, rot, disp_intensity, judge,
trial_time)
event.waitKeys(keyList=[
thiskey
]) # press the response key again to start the next trial
config_tools.write_xrl(self.subject, xls_file=xlsname)
# save results in xls-file
workbook = xlsxwriter.Workbook(xlsname)
for handler_idx, cur_handler in enumerate(stairs):
worksheet = workbook.add_worksheet(
cur_handler.extraInfo['label'])
worksheet.write('A1', 'Reversal Intensities')
worksheet.write('B1', 'Reversal Indices')
worksheet.write('C1', 'All Intensities')
worksheet.write('D1', 'All Responses')
for i in range(len(rot_all[handler_idx])):
# worksheet.write('C' + str(i + 2), rot_all[handler_idx][i])
worksheet.write('C' + str(i + 2),
rot_all_disp[handler_idx][i])
worksheet.write('D' + str(i + 2),
judge_all[handler_idx][i])
workbook.close()
# print resulting parameters and estimates for each step
res_file_path = os.path.join(path, self.idx + '_estimates.csv')
res_writer = csv.writer(open(res_file_path, 'w'))
for res_stim, res_vals in estimates.items():
for res_val_id, res_val in enumerate(res_vals):
res_writer.writerow([
res_stim, res_val_id, res_val[0], res_val[1],
res_val[2]
])
# save each handler into a psydat-file and save posterior into a numpy-file
for cur_handler in stairs:
file_name = os.path.join(
path, self.idx + self.param['noise_condition'] +
cur_handler.extraInfo['label'])
misc.toFile(file_name + '.psydat', cur_handler)
if isinstance(cur_handler, data.PsiHandler):
cur_handler.savePosterior(file_name + '.npy')
thisStair.addResponse(correct, isoi)
trialData.write('{},{},{},{}\n' .format(thisStair.extraInfo['Label'],isoi, direction, int(correct)))
print('{} of {} trials complete\n' .format(trialNum+1, nTrials))
print('\n=== EXPERIMENT FINISHED ===\n')
## ----
## -- use quest to estimate threshold --
if exptInfo['09. Number of trials per staircase'] > 0:
for s in staircases:
s.saveAsPickle(fileName+'_'+s.extraInfo['Label']) #special python binary file to save all the info
q = data.QuestHandler(log(exptInfo['06. First ISOI (ms)']),
log(exptInfo['06. First ISOI (ms)']),
pThreshold = 0.82,
nTrials = exptInfo['09. Number of trials per staircase'],
grain=0.1,
range=10,
minVal = log(exptInfo['11. Min ISOI (ms)']),
maxVal = log(exptInfo['12. Max ISOI (ms)']))
isois = [log(i) for i in s.intensities]
q.importData(isois,s.data)
threshold = exp(q.mean())
tSD = exp(q.sd())
print('\n'+s.extraInfo['Label'])
print('threshold: {}' .format(threshold))
print('sd: {}' .format(tSD))
thresholdData.write('{},{},{},{},{},{},{},{},{},{}\n' .format(threshold,tSD,
s.extraInfo['Label'],
s.extraInfo['direction'],
exptInfo['02. Test number'],
def test_json_dump_to_file(self):
q = data.QuestHandler(0.5, 0.2, pThreshold=0.63, gamma=0.01,
nTrials=20, minVal=0, maxVal=1)
q.saveAsJson(fileName=self.tmp_dir, fileCollisionMethod='overwrite')
def run_session(self):
path = os.path.join(self.res_dir, self.subject)
if not os.path.exists(path):
os.makedirs(path)
psydat_path = os.path.join(path, 'psydat')
if not os.path.exists(psydat_path):
os.makedirs(psydat_path)
# welcome
msg = visual.TextStim(self.win,
'Welcome!' + '\n' +
' Press any key to start this session :)',
color='black',
units='deg',
pos=(0, 0),
height=0.8)
msg.draw()
self.win.mouseVisible = False
self.win.flip()
event.waitKeys()
# read staircase parameters
conditions = [
dict({'stimulus': key}, **value)
for key, value in self.param.items() if key.startswith('stimulus')
]
if conditions[0]['stairType'] == 'simple':
stairs = data.MultiStairHandler(stairType='simple',
conditions=conditions,
nTrials=self.trial_nmb,
method='sequential')
elif conditions[0]['stairType'] == 'quest':
stairs = []
for cond in conditions:
if self.priors_file_path:
prior_file = self.priors_file_path + cond[
'label'] + '.psydat'
print(prior_file)
prior_handler = misc.fromFile(prior_file)
else:
prior_handler = None
cur_handler = data.QuestHandler(cond['startVal'],
cond['startValSd'],
pThreshold=cond['pThreshold'],
nTrials=self.trial_nmb,
minVal=cond['min_val'],
maxVal=cond['max_val'],
staircase=prior_handler,
extraInfo=cond,
grain=0.02)
stairs.append(cur_handler)
elif conditions[0]['stairType'] == 'psi':
stairs = []
for cond in conditions:
if self.priors_file_path:
prior_file = self.priors_file_path + cond['label'] + '.npy'
else:
prior_file = None
print(prior_file)
cur_handler = data.PsiHandler(nTrials=self.trial_nmb,
intensRange=[1, 10],
alphaRange=[1, 10],
betaRange=[0.01, 10],
intensPrecision=0.1,
alphaPrecision=0.1,
betaPrecision=0.01,
delta=0.01,
extraInfo=cond,
prior=prior_file,
fromFile=(prior_file
is not None))
stairs.append(cur_handler)
# write configuration files
xpp = config_tools.WriteXpp(self.subject, self.idx)
xpp_file = xpp.head(self.cfg_file, self.par_file)
config_tools.write_xrl(self.subject,
cfg_file=self.cfg_file,
par_file=self.par_file,
xpp_file=xpp_file)
xlsname = path + '/' + self.idx + self.param[
'noise_condition'] + '.xlsx'
""" running staircase """
if isinstance(stairs, data.MultiStairHandler):
# start running the staircase using the MultiStairHandler for the up-down method
count = 0
for rot, cond in stairs:
count += 1
direction = (-1)**(cond['label'].endswith('m')
) # direction as -1 if for minus stim
rot = rot * direction # rotation for this trial
judge, react_time, trial_time_start = self.run_trial(
rot, cond, cond['std'], count)
# check whether the theta is valid - if not, the rotation given by staircase should be corrected by
# realizable values
valid_theta = np.round(np.load(self.hue_list), decimals=1)
disp_standard = self.take_closest(
valid_theta, cond['standard']) # theta actually displayed
stair_test = cond[
'standard'] + stairs._nextIntensity * direction
if stair_test < 0:
stair_test += 360
disp_test = self.take_closest(valid_theta, stair_test)
disp_intensity = disp_test - disp_standard
if disp_intensity > 300:
disp_intensity = (disp_test + disp_standard) - 360
stairs.addResponse(judge, abs(disp_intensity))
xpp.task(count, cond, rot, float(disp_intensity), judge,
react_time, trial_time_start)
if 'escape' in event.waitKeys():
config_tools.write_xrl(self.subject,
break_info='userbreak')
core.quit()
config_tools.write_xrl(self.subject, xls_file=xlsname)
stairs.saveAsExcel(xlsname) # save results
misc.toFile(
os.path.join(
psydat_path,
self.idx + self.param['noise_condition'] + '.psydat'),
stairs)
elif isinstance(stairs, list):
# start running the staircase using custom interleaving stairs for the quest and psi methods
count = 0
rot_all = []
rot_all_disp = []
judge_all = []
estimates = {s.extraInfo['label']: [] for s in stairs}
for trial_n in range(self.trial_nmb):
for handler_idx, cur_handler in enumerate(stairs):
count += 1
direction = (-1)**(
cur_handler.extraInfo['label'].endswith('m')
) # direction as -1 if for minus stim
if cur_handler._nextIntensity >= 10.0:
sys.exit(
"Hue difference is out of range! Please enlarge the testing range or take more training!"
)
rot = cur_handler._nextIntensity * direction # rotation for this trial
if trial_n >= 5: # avoid repeating an intensity more than 3 times
last_rots = [
np.round(r, decimals=1) for r in [
rot_all_disp[handler_idx][trial_n - 1],
rot_all_disp[handler_idx][trial_n - 2],
rot_all_disp[handler_idx][trial_n - 3]
]
]
last_resp = [
judge_all[handler_idx][trial_n - 1],
judge_all[handler_idx][trial_n - 2],
judge_all[handler_idx][trial_n - 3]
]
if last_rots[0] == last_rots[1] == last_rots[2] \
and last_resp[0] == last_resp[1] == last_resp[2]:
if cur_handler._nextIntensity > 0.5:
rot = (cur_handler._nextIntensity -
0.5) * direction
print('Intensity decreases by 0.5!')
if cur_handler._nextIntensity <= 0.5:
rot = (cur_handler._nextIntensity +
0.5) * direction
print('Intensity increases by 0.5!')
cond = cur_handler.extraInfo
judge, react_time, trial_time_start = self.run_trial(
rot, cond, cond['std'], count)
if len(rot_all) <= handler_idx:
rot_all.append([])
rot_all[handler_idx].append(rot)
if len(judge_all) <= handler_idx:
judge_all.append([])
judge_all[handler_idx].append(judge)
valid_theta = np.round(np.load(self.hue_list), decimals=1)
disp_standard = self.take_closest(valid_theta,
cond['standard'])
stair_test = cond[
'standard'] + rot # calculated test hue for this trial
if stair_test < 0:
stair_test += 360
disp_test = self.take_closest(
valid_theta,
stair_test) # actual displayed test hue for this trial
disp_intensity = disp_test - disp_standard # actual displayed hue difference
if disp_intensity > 300:
disp_intensity = (disp_test + disp_standard) - 360
cur_handler.addResponse(
judge, abs(disp_intensity)
) # only positive number is accepted by addResponse
if len(rot_all_disp
) <= handler_idx: # add displayed intensities
rot_all_disp.append([])
rot_all_disp[handler_idx].append(disp_intensity)
if isinstance(cur_handler, data.PsiHandler):
estimates[cur_handler.extraInfo['label']].append([
cur_handler.estimateLambda()[0], # location
cur_handler.estimateLambda768()[1], # slope
cur_handler.estimateThreshold(0.75)
])
elif isinstance(cur_handler, data.QuestHandler):
estimates[cur_handler.extraInfo['label']].append([
cur_handler.mean(),
cur_handler.mode(),
cur_handler.quantile(0.5)
])
xpp.task(count, cond, rot, disp_intensity, judge,
react_time, trial_time_start)
if 'escape' in event.waitKeys():
config_tools.write_xrl(self.subject,
break_info='userbreak')
core.quit()
config_tools.write_xrl(self.subject, xls_file=xlsname)
# save results in xls-file
workbook = xlsxwriter.Workbook(xlsname)
for handler_idx, cur_handler in enumerate(stairs):
worksheet = workbook.add_worksheet(
cur_handler.extraInfo['label'])
worksheet.write('A1', 'Reversal Intensities')
worksheet.write('B1', 'Reversal Indices')
worksheet.write('C1', 'All Intensities')
worksheet.write('D1', 'All Responses')
for i in range(len(rot_all[handler_idx])):
# worksheet.write('C' + str(i + 2), rot_all[handler_idx][i])
worksheet.write('C' + str(i + 2),
rot_all_disp[handler_idx][i])
worksheet.write('D' + str(i + 2),
judge_all[handler_idx][i])
workbook.close()
# print resulting parameters and estimates for each step
res_file_path = os.path.join(path, self.idx + '_estimates.csv')
res_writer = csv.writer(open(res_file_path, 'w'))
for res_stim, res_vals in estimates.items():
for res_val_id, res_val in enumerate(res_vals):
res_writer.writerow([
res_stim, res_val_id, res_val[0], res_val[1],
res_val[2]
])
# save each handler into a psydat-file and save posterior into a numpy-file
for cur_handler in stairs:
file_name = os.path.join(
psydat_path, self.idx + self.param['noise_condition'] +
cur_handler.extraInfo['label'])
misc.toFile(file_name + '.psydat', cur_handler)
if isinstance(cur_handler, data.PsiHandler):
cur_handler.savePosterior(file_name + '.npy')
ax2.tick_params(axis='x',which='minor',bottom='off')
# #save figure to file
# outputFile = os.path.join(dataDir, 'test.pdf')
# pylab.savefig(outputFile)
if __name__ == "__main__":
#Test staircase functions
threshCriterion = 0.25
staircaseTrials = 5
staircase = data.QuestHandler(startVal = 95,
startValSd = 80,
stopInterval= 1, #sd of posterior has to be this small or smaller for staircase to stop, unless nTrials reached
nTrials = staircaseTrials,
#extraInfo = thisInfo,
pThreshold = threshCriterion, #0.25,
gamma = 1./26,
delta=0.02, #lapse rate, I suppose for Weibull function fit
method = 'quantile', #uses the median of the posterior as the final answer
stepType = 'log', #will home in on the 80% threshold. But stepType = 'log' doesn't usually work
minVal=1, maxVal = 100
)
print('created QUEST staircase')
descendingPsycho = False
noiseEachTrial = np.array([5,5,5,5,5,5,5,5,5,10,10,10,10,10,10,10,10,10,10,10,20,20,20,20,20,20,20,20,20,20,20,20,50,50,50,50,50,50,50,60,60,60,60,60,60,60,60,60,60,70,70,70,70,70,70,70,80,80,80,80,80,80,80,80,95,95,95,95,95,95,95])
centeredOnZero = noiseEachTrial/100. -0.5
guessRate = .1 #doesnt work with guessRate=0, fitWeibull doesnt like that
pCorrEachTrial = guessRate*.5 + (1-guessRate)* 1. / (1. + np.exp(-20*centeredOnZero)) #sigmoidal probability
print('pCorrEachTrial=',np.around(pCorrEachTrial,2))
corrEachTrial = np.zeros( len(noiseEachTrial) )
screen.setSizePix([1680, 1050])
screen.setWidth(47.475)
screen.setDistance(57)
win = visual.Window(allowGUI=True, units='deg', monitor=screen)
grating = visual.GratingStim(win,
tex='sin',
mask='gauss',
contrast=1,
sf=3,
size=3)
staircase = data.QuestHandler(0.5,
0.2,
pThreshold=0.63,
gamma=0.01,
minVal=0,
maxVal=1,
ntrials=10)
orientations = [-45, 45]
responses = ['left', 'right']
for contrast in staircase:
keys = []
# randomise orientation for this trial
grating.ori = np.random.choice(orientations)
# update the difficulty (the contrast)
grating.contrast = contrast
# before the trial: wait 500ms
core.wait(0.5)
for thisComponent in instrPracticeComponents:
if hasattr(thisComponent, "setAutoDraw"):
thisComponent.setAutoDraw(False)
# check responses
if ok1.keys in ['', [], None]: # No response was made
ok1.keys=None
thisExp.nextEntry()
# the Routine "instrPractice" was not non-slip safe, so reset the non-slip timer
routineTimer.reset()
# --------Prepare to start Staircase "trials" --------
# set up handler to look after next chosen value etc
if expInfo['participant'] == 'ideal' :
trials = data.QuestHandler(startVal = 0.1,
startValSd = 0.05, delta=0,
pThreshold = 0.63,
nTrials=1E3, minVal=0, maxVal=1)
else :
trials = data.QuestHandler(startVal = 0.5,
startValSd = 0.2,
pThreshold = 0.63,
nTrials=50, minVal=0, maxVal=1)
thisExp.addLoop(trials) # add the loop to the experiment
level = thisTrial = 0.5 # initialise some vals
face = visual.ImageStim(win, happy_face)
key_resp_2 = event.BuilderKeyResponse()
for thisTrial in trials:
currentLoop = trials
level = thisTrial
def runQuest(experimentStructure, stimuliStructure, data, feedbackText,
pThreshold, questTrials):
""" This function implements a Quest procedure to individualize the maximum contrast difference
Currently not in use.
Input:
exerimentStructure: all general experimental properties
stimuliStructure: all general stimulus properties
data: PsychoPy data functions
feedbackText: feedback text object instance
pThreshold: threshold for quest procedure
questTrials: number of trials for qQest procedure
Return:
trials.quantile: quantile of Quest posterior pdf
accPerf: accumulated performance of current block
"""
# Create some shortnames
thisExp = experimentStructure['thisExp']
patch1 = stimuliStructure['patch1']
patch2 = stimuliStructure['patch2']
questClock = stimuliStructure['questClock']
# Create some info to store with the data
info = {}
info['startPoints'] = [0.03, 0.04]
info['nTrials'] = questTrials
lowerBound = 0.01
upperBound = 0.05
whichLoop = 'quest'
stairs = []
for thisStart in info['startPoints']:
# We need a COPY of the info for each staircase
# (or the changes here will be made to all the other staircases)
thisInfo = copy.copy(info)
# Now add any specific info for this staircase
thisInfo['thisStart'] = thisStart #we might want to keep track of this
trials = data.QuestHandler(startVal=thisStart,
startValSd=0.1,
extraInfo=thisInfo,
pThreshold=pThreshold,
nTrials=questTrials,
minVal=lowerBound,
maxVal=upperBound,
name='quest',
gamma=0.5)
thisExp.addLoop(trials)
stairs.append(trials)
accPerf = 0.0
for trialN in range(info['nTrials']):
for trials in stairs:
thisDifference = trials.next()
print 'start=%.2f, current=%.4f' % (trials.extraInfo['thisStart'],
thisDifference)
lowerBound = float('NaN')
upperBound = float('NaN')
positions = np.random.choice([0, 1])
if positions == 1:
targetPatch = 'left'
else:
targetPatch = 'right'
(decision1, trials) = createPatches(experimentStructure,
stimuliStructure, questClock,
lowerBound, upperBound,
targetPatch, trials, whichLoop,
thisDifference)
if (decision1.keys == None):
msg = 'Zu langsam!'
giveFeedback(experimentStructure, stimuliStructure,
feedbackText, msg)
if decision1.corr:
accPerf = accPerf + 1
thisExp.nextEntry()
questQuantiles = []
for trials in stairs:
questQuantiles.append(trials.quantile())
print(questQuantiles)
print(np.mean(questQuantiles))
return (trials.quantile(), accPerf)
def main():
# Sets system time as the random seed
np.random.seed(seed=None)
# Prompt GUI for participant ID
id = ask_for_id()
# Set up window and how many frames image should be displayed depending on refresh rate of monitor
window = visual.Window(color=bg_color,
monitor='monitor',
fullscr=full_screen)
frame_rate = window.getActualFrameRate()
# Display instructional messages for user
instruction_msg = visual.TextStim(
window,
color=text_color,
text=
f"For each image, answer whether the building shown is damaged. Each image will be shown briefly.\n\nTo answer, please press {key_list[0]} for yes and {key_list[1]} for no.\n\nPress any key to continue."
)
instruction_msg.draw()
window.flip()
event.waitKeys()
instruction_msg.text = f"For the first few images, you will hear a beep if you give the correct answer.\n\nPress any key to continue."
instruction_msg.draw()
window.flip()
event.waitKeys()
plus_horiz = visual.ShapeStim(window,
units='pix',
vertices=[[-20, 0], [20, 0]])
plus_vert = visual.ShapeStim(window,
units='pix',
vertices=[[0, 20], [0, -20]])
user_data = pd.DataFrame(columns=data_columns)
img = visual.ImageStim(window, size=img_dims)
# Setup staircase procedures using defined parameters
conds = data.importConditions('demo_params.csv')
staircases = []
for cond in conds:
staircases.append(
data.QuestHandler(startVal=cond['startVal'],
startValSd=cond['startValSd'],
pThreshold=cond['pThreshold'],
nTrials=cond['nTrials'],
beta=cond['beta'],
delta=cond['delta'],
gamma=cond['gamma'],
minVal=cond['minVal'],
maxVal=cond['maxVal']))
img_list = get_imgs(img_dir, damage_subdir, num_each, nodamage_subdir)
print(f"{len(img_list)} images loaded.")
key_list.append('escape')
# Run through image list with participant
rounds = 1
while len(img_list) > 0:
staircase = random.choice(staircases)
curr_time = next(staircase)
curr_time = float(format(curr_time, '.3f'))
# Get and parse random image's information
subdir, img_name = get_random_img(img_list)
truth = 'y' if subdir == damage_subdir else 'n'
img.setImage(img_dir + subdir + '/' + img_name)
# Display the image for set number of frames
frames_to_show = get_frames_to_show(curr_time, frame_rate)
keypress = None
start_time, end_time = time.time(), 0
for _ in range(frames_to_show):
img.draw()
window.flip()
keypress = event.getKeys(keyList=key_list)
if 'escape' in keypress:
instruction_msg.text = 'Experiment aborted. Quitting...'
instruction_msg.draw()
window.flip()
core.wait(3.0)
core.quit()
if len(keypress) > 0:
end_time = time.time()
break
plus_horiz.draw()
plus_vert.draw()
window.flip()
# Track reaction time for user response
if keypress is None or len(keypress) == 0:
keypress = event.waitKeys(keyList=key_list)
if 'escape' in keypress:
instruction_msg.text = 'Experiment aborted. Quitting...'
instruction_msg.draw()
window.flip()
core.wait(3.0)
core.quit()
end_time = time.time()
reaction_time = float(format(end_time - start_time, '.3f'))
# Log and process user's input as correct or incorrect
answer = 'y' if keypress[0] == key_list[0] else 'n'
event.clearEvents()
if answer == truth:
if rounds <= feedback_rounds:
beep = sound.Sound('A', secs=0.25)
beep.play()
staircase.addResponse(1)
else:
staircase.addResponse(0)
user_data.loc[len(user_data)] = ([
img_name, answer, truth, curr_time, reaction_time
])
rounds += 1
if rounds == feedback_rounds + 1:
instruction_msg.text = 'From now on, feedback will no longer be provided.'
instruction_msg.draw()
window.flip()
core.wait(3.0)
# Calculate the final presentation time from the average of the staircases
avg = 0
for staircase in staircases:
avg += next(staircase)
avg /= len(staircases)
with open('demo_result.txt', 'w') as f:
f.write(f"The final presentation time is {avg} seconds.\n")
user_data['Correct'] = (
user_data['Actual'] == user_data['Response']).astype(int)
# Output individual participant data to .csv file
if not os.path.exists(data_dir):
os.makedirs(data_dir)
user_data.to_csv(f"{data_dir}/data_{id}.csv")
instruction_msg.text = "Test completed. Closing window..."
instruction_msg.draw()
window.flip()
core.wait(3.0)
core.quit()
'observer': subject_initials,
}
expInfo['dateStr'] = data.getDateStr() # add the current time
# make a text file to save data
fileName = expInfo['observer'] + expInfo['dateStr']
dataFile = open(fileName + '.csv',
'w') # a simple text file with 'comma-separated-values'
dataFile.write('targetSide,volume,correct\n')
# create the staircase handler
staircase = data.QuestHandler(startVal=0.5,
startValSd=0.05,
pThreshold=0.65,
gamma=0.5,
beta=3.5,
delta=0.01,
nTrials=nr_trials,
minVal=0,
maxVal=1)
# create window and stimuli
win = visual.Window([1920, 1080],
fullscr=True,
allowGUI=True,
monitor='testMonitor',
units='pixels')
fixation = visual.GratingStim(win,
pos=(0, 0),
tex='sin',
mask='circle',
gen_concentration_steps()[exp_info['Substance']])
start_val = np.log10(get_start_val(exp_info['Substance']))
quest_params = dict(
startVal=start_val,
startValSd=np.log10(20),
pThreshold=0.82,
beta=3.5,
gamma=0.01,
delta=0.01,
grain=0.01,
range=2 * np.abs(concentration_steps.max() - concentration_steps.min()),
nTrials=20,
stopInterval=None)
quest_handler = data.QuestHandler(**quest_params)
exp_handler = data.ExperimentHandler(extraInfo=exp_info,
savePickle=True,
saveWideText=True,
dataFileName=os.path.join(
path['rawdata_dir'],
path['base_filename']))
exp_handler.addLoop(quest_handler)
for concentration_proposed in quest_handler:
trial_number = quest_handler.thisTrialN + 1 # QuestHandler counts from 0.
# QUEST proposes a concentration to present,
# but we usually don't have that particular one prepared.
#
# So we pick the one from our set which is closest to the proposed.