def make(cls, **kwargs):
seed = kwargs.get('seed')
prng = random.RandomState(seed)
ratio_cue_valid = kwargs.get('ratio_cue_valid', 0.67)
trials = pandas.DataFrame({'cue_type': ['arrow', 'word']})
trials = expand(trials, 'cue_validity', values=['valid', 'invalid'],
ratio=ratio_cue_valid, seed=seed)
trials = extend(trials, max_length=320)
trials['target_loc'] = prng.choice(['left', 'right'], len(trials))
trials['correct_response'] = trials['target_loc']
reverser = dict(left='right', right='left')
def pick_cue_dir(trial):
cue_validity = trial['cue_validity']
target_loc = trial['target_loc']
if cue_validity == 'valid':
return target_loc
elif cue_validity == 'invalid':
return reverser[target_loc]
else:
raise NotImplementedError('cue_validity: %s' % cue_validity)
trials['cue_dir'] = trials.apply(pick_cue_dir, axis=1)
# Determine cue pos and target pos
trials['cue_pos_dy'] = 0.
trials['target_pos_dy'] = 0.
trials[['cue_pos_dy', 'target_pos_dy']] = prng.multivariate_normal(
mean=[0, 0], cov=[[1, 0], [0, 1]], size=len(trials)
)
# Actual units (pixels) are determined at runtime.
for x in ['target_pos_x', 'target_pos_y', 'cue_pos_y']:
trials[x] = ''
trials = add_block(trials, size=60, start=1, seed=seed)
trials['block_type'] = 'test'
# Add practice trials
num_practice = 8
practice_ix = prng.choice(trials.index, num_practice)
practice_trials = trials.ix[practice_ix, ]
trials.drop(practice_ix, inplace=True)
practice_trials['block'] = 0
practice_trials['block_type'] = 'practice'
trials = pandas.concat([practice_trials, trials])
trials['trial'] = range(len(trials))
# Add blank columns for response variables
for x in ['response', 'rt', 'is_correct']:
trials[x] = ''
return cls(trials.to_dict('record'))
def make(cls, **kwargs):
"""Create a list of trials.
Each trial is a dict with values for all keys in self.COLUMNS.
"""
# make convenience copies of class defaults
stim_dir = cls.STIM_DIR
columns = cls.COLUMNS
settings = kwargs.copy()
seed = settings.get('seed')
try:
seed = int(seed)
except (TypeError, ValueError):
logging.info('seed was None or couldn\'t be converted to int')
seed = None
prng = random.RandomState(seed)
# Read in the trials (same for all participants)
trials = pd.read_csv(Path(stim_dir, 'propositions.csv'))
# Randomly assign mask trials
trials['mask_type'] = prng.choice(['mask', 'nomask'], size=len(trials))
# Choose a version of each cue at random
all_cues = Path(stim_dir, 'cues').listdir('*.wav', names_only=True)
def pick_cue_file(cue):
options = [c for c in all_cues if c.find(cue) == 0]
return prng.choice(options)
trials['cue_file'] = trials.cue.apply(pick_cue_file)
# Add columns for response variables
for col in ['response', 'rt', 'is_correct']:
trials[col] = ''
# Add practice trials
num_practice = 6
practice_ix = prng.choice(trials.index, num_practice, replace=False)
practice_trials = trials.ix[practice_ix, ]
practice_trials['block'] = 0
practice_trials['block_type'] = 'practice'
trials.drop(practice_ix, inplace=True)
# Finishing touches: block trials and shuffle
num_blocks = 4
trials_per_block = (len(trials)/num_blocks) + 1
trials = add_block(trials, trials_per_block, name='block', start=1,
groupby='cue', seed=seed)
trials = smart_shuffle(trials, col='cue', block='block', seed=seed)
trials['block_type'] = 'test'
# Merge practice trials
trials = pd.concat([practice_trials, trials], ignore_index=True)
# Label trial
trials['trial'] = range(1, len(trials)+1)
# Reorder columns
trials = trials[columns]
return cls(trials.to_dict('record'))
def spatial_cueing_trial_list(cue_type, mask_type, **participant_kwargs):
trials = counterbalance({
'target_loc': ['left', 'right'],
'cue_type': cue_type,
'mask_type': mask_type,
})
seed = participant_kwargs.get('seed')
random.seed(seed)
# Determine cue validity, starting with the smallest group
# 66.6% valid, 25% invalid, 8.3% neutral
trials = expand(trials,
'cue_validity',
values=['invalid', 'neutral'],
ratio=0.75,
seed=seed)
trials = expand(trials,
'tmp_cue_valid',
values=[1, 0],
ratio=0.7,
seed=seed)
trials.loc[trials.tmp_cue_valid == 1, 'cue_validity'] = 'valid'
del trials['tmp_cue_valid']
# Take anything given by the participant and put it in the trial list
for col_name, subj_info in participant_kwargs.items():
trials[col_name] = subj_info
# Target was determined, now determine cue dir based on cue validity
direction_reverser = {'left': 'right', 'right': 'left'}
def determine_cue_dir(trial):
target_loc = trial['target_loc']
if trial['cue_validity'] in 'valid':
return target_loc
elif trial['cue_validity'] == 'invalid':
return direction_reverser[target_loc]
elif trial['cue_validity'] == 'neutral':
return 'neutral'
else:
msg = 'cue validity %s not implemented' % trial['cue_validity']
raise NotImplementedError(msg)
trials['cue_dir'] = trials.apply(determine_cue_dir, axis=1)
# Save a copy of the practice trials
practice_trials = trials.copy()
practice_trials['block'] = 0
# Ensure that there are some up/down targets in the practice trials
practice_trial_invalid_ix = practice_trials.ix[
practice_trials.cue_validity == 'invalid', ].index
practice_trial_target_loc = [
random.choice(['up', 'down']) for _ in practice_trial_invalid_ix
]
practice_trials.ix[practice_trial_invalid_ix, 'target_loc'] = \
practice_trial_target_loc
# Select a subset of trials for practice
num_practice_trials = 15
sampled_trials = random.sample(practice_trials.index, num_practice_trials)
practice_trials = practice_trials.ix[sampled_trials]
practice_trials.reset_index(drop=True)
# Duplicate unique trials evenly to reach max
# 320 trials ~ 20 trials in each within subject cell
trials = extend(trials, max_length=320)
# Set target location for catch trials
invalid_trial_ix = trials.ix[trials.cue_validity == 'invalid', ].index
catch_trial_ix = random.sample(invalid_trial_ix, len(invalid_trial_ix) / 2)
catch_trial_target_loc = [
random.choice(['up', 'down']) for _ in catch_trial_ix
]
trials.ix[catch_trial_ix, 'target_loc'] = catch_trial_target_loc
# Assign block randomly
block_size = 80
trials = add_block(trials,
size=block_size,
id_col='cue_validity',
start_at=1,
seed=seed)
# Join the practice trials
trials = pandas.concat([practice_trials, trials])
# Shuffle by block
trials = simple_shuffle(trials, block='block', seed=seed)
trials = trials.reset_index(drop=True)
# Label trial number
trials.insert(0, 'trial', range(len(trials)))
# Rearrange columns
participant_keys = [
'subj_id',
'seed',
'sona_experiment_code',
'experimenter',
'cue_contrast',
]
for p in participant_keys:
if p not in trials.columns:
trials[p] = ''
col_order = participant_keys + [
'block',
'trial',
'mask_type',
'cue_type',
'cue_validity',
'cue_dir',
'target_loc',
]
assert all([c in trials.columns for c in col_order])
trials = trials[col_order]
# Fill columns decided at runtime
trials['soa'] = ''
trials['target_loc_x'] = ''
trials['target_loc_y'] = ''
# Fill response columns
trials['rt'] = ''
trials['response_type'] = ''
trials['is_correct'] = ''
return trials
def make(cls, **kwargs):
""" Create a list of trials.
Each trial is a dict with values for all keys in self.COLUMNS.
"""
settings = dict(cls.DEFAULTS)
settings.update(kwargs)
seed = settings.get('seed')
prng = pd.np.random.RandomState(seed)
# Balance within subject variables
trials = counterbalance({'feat_type': ['visual', 'nonvisual'],
'mask_type': ['mask', 'nomask']})
trials = expand(trials, name='correct_response', values=['yes', 'no'],
ratio=settings['ratio_yes_correct_responses'],
seed=seed)
trials = expand(trials, name='response_type', values=['prompt', 'pic'],
ratio=settings['ratio_prompt_response_type'],
seed=seed)
# Extend the trials to final length
trials = extend(trials, reps=4)
# Read proposition info
propositions_csv = Path(cls.STIM_DIR, 'propositions.csv')
propositions = pd.read_csv(propositions_csv)
# Add cue
categories = propositions.cue.unique()
trials['cue'] = prng.choice(categories, len(trials), replace=True)
_propositions = propositions.copy()
def determine_question(row):
is_cue = (_propositions.cue == row['cue'])
is_feat_type = (_propositions.feat_type == row['feat_type'])
is_correct_response = (_propositions.correct_response ==
row['correct_response'])
valid_propositions = (is_cue & is_feat_type & is_correct_response)
if valid_propositions.sum() == 0:
trials.ix[row.name, 'cue'] = prng.choice(categories)
return determine_question(trials.ix[row.name, ])
options = _propositions.ix[valid_propositions, ]
selected_ix = prng.choice(options.index)
selected_proposition_id = options.ix[selected_ix, 'proposition_id']
_propositions.drop(selected_ix, inplace=True)
return selected_proposition_id
trials['proposition_id'] = trials.apply(determine_question, axis=1)
# Merge in question
trials = trials.merge(propositions)
# Add in picture
def determine_pic(row):
if row['response_type'] != 'pic':
return ''
elif row['correct_response'] == 'yes':
return row['cue']
else:
distractors = list(categories)
distractors.remove(row['cue'])
return prng.choice(distractors)
trials['pic'] = trials.apply(determine_pic, axis=1)
# Add columns for response variables
for col in ['response', 'rt', 'is_correct']:
trials[col] = ''
# Add practice trials
num_practice = 8
practice_ix = prng.choice(trials.index, num_practice)
practice_trials = trials.ix[practice_ix, ]
practice_trials['block'] = 0
practice_trials['block_type'] = 'practice'
trials.drop(practice_ix, inplace=True)
# Finishing touches
trials = add_block(trials, 50, name='block', start=1, groupby='cue',
seed=seed)
trials = smart_shuffle(trials, col='cue', block='block', seed=seed)
trials['block_type'] = 'test'
# Merge practice trials
trials = pd.concat([practice_trials, trials])
# Label trial
trials['trial'] = range(len(trials))
# Reorcder columns
trials = trials[cls.COLUMNS]
return cls(trials.to_dict('record'))
def make(cls, **kwargs):
""" Create a list of trials.
Each trial is a dict with values for all keys in self.COLUMNS.
"""
settings = dict(cls.DEFAULTS)
settings.update(kwargs)
seed = settings.get('seed')
prng = pd.np.random.RandomState(seed)
# Balance within subject variables
trials = counterbalance({
'feat_type': ['visual', 'nonvisual'],
'mask_type': ['mask', 'nomask']
})
trials = expand(trials,
name='correct_response',
values=['yes', 'no'],
ratio=settings['ratio_yes_correct_responses'],
seed=seed)
trials = expand(trials,
name='response_type',
values=['prompt', 'pic'],
ratio=settings['ratio_prompt_response_type'],
seed=seed)
# Extend the trials to final length
trials = extend(trials, reps=4)
# Read proposition info
propositions_csv = Path(cls.STIM_DIR, 'propositions.csv')
propositions = pd.read_csv(propositions_csv)
# Add cue
categories = propositions.cue.unique()
trials['cue'] = prng.choice(categories, len(trials), replace=True)
_propositions = propositions.copy()
def determine_question(row):
is_cue = (_propositions.cue == row['cue'])
is_feat_type = (_propositions.feat_type == row['feat_type'])
is_correct_response = (
_propositions.correct_response == row['correct_response'])
valid_propositions = (is_cue & is_feat_type & is_correct_response)
if valid_propositions.sum() == 0:
trials.ix[row.name, 'cue'] = prng.choice(categories)
return determine_question(trials.ix[row.name, ])
options = _propositions.ix[valid_propositions, ]
selected_ix = prng.choice(options.index)
selected_proposition_id = options.ix[selected_ix, 'proposition_id']
_propositions.drop(selected_ix, inplace=True)
return selected_proposition_id
trials['proposition_id'] = trials.apply(determine_question, axis=1)
# Merge in question
trials = trials.merge(propositions)
# Add in picture
def determine_pic(row):
if row['response_type'] != 'pic':
return ''
elif row['correct_response'] == 'yes':
return row['cue']
else:
distractors = list(categories)
distractors.remove(row['cue'])
return prng.choice(distractors)
trials['pic'] = trials.apply(determine_pic, axis=1)
# Add columns for response variables
for col in ['response', 'rt', 'is_correct']:
trials[col] = ''
# Add practice trials
num_practice = 8
practice_ix = prng.choice(trials.index, num_practice)
practice_trials = trials.ix[practice_ix, ]
practice_trials['block'] = 0
practice_trials['block_type'] = 'practice'
trials.drop(practice_ix, inplace=True)
# Finishing touches
trials = add_block(trials,
50,
name='block',
start=1,
groupby='cue',
seed=seed)
trials = smart_shuffle(trials, col='cue', block='block', seed=seed)
trials['block_type'] = 'test'
# Merge practice trials
trials = pd.concat([practice_trials, trials])
# Label trial
trials['trial'] = range(len(trials))
# Reorcder columns
trials = trials[cls.COLUMNS]
return cls(trials.to_dict('record'))
