def test_rt_chisq():
"""Test reaction time chi-square fitting
"""
# 1D should return single float
foo = np.random.rand(30)
assert_raises(ValueError, ea.rt_chisq, foo - 1.)
assert_equal(np.array(ea.rt_chisq(foo)).shape, ())
# 2D should return array with shape of input but without ``axis`` dimension
foo = np.random.rand(30).reshape((2, 3, 5))
for axis in range(-1, foo.ndim):
bar = ea.rt_chisq(foo, axis=axis)
assert_true(np.all(np.equal(np.delete(foo.shape, axis),
np.array(bar.shape))))
def test_rt_chisq():
"""Test reaction time chi-square fitting."""
# 1D should return single float
foo = np.random.RandomState(0).rand(30)
pytest.raises(ValueError, ea.rt_chisq, foo - 1.)
assert_equal(np.array(ea.rt_chisq(foo, warn=False)).shape, ())
# 2D should return array with shape of input but without ``axis`` dimension
foo = np.random.rand(30).reshape((2, 3, 5))
for axis in range(-1, foo.ndim):
bar = ea.rt_chisq(foo, axis=axis, warn=False)
assert_array_equal(np.delete(foo.shape, axis), np.array(bar.shape))
foo_bad = np.concatenate((np.random.rand(30), [100.]))
with pytest.warns(UserWarning, match='likely bad'):
bar = ea.rt_chisq(foo_bad)
def test_rt_chisq():
"""Test reaction time chi-square fitting."""
# 1D should return single float
foo = np.random.RandomState(0).rand(30)
pytest.raises(ValueError, ea.rt_chisq, foo - 1.)
assert_equal(np.array(ea.rt_chisq(foo, warn=False)).shape, ())
# 2D should return array with shape of input but without ``axis`` dimension
foo = np.random.rand(30).reshape((2, 3, 5))
for axis in range(-1, foo.ndim):
bar = ea.rt_chisq(foo, axis=axis, warn=False)
assert_array_equal(np.delete(foo.shape, axis), np.array(bar.shape))
foo_bad = np.concatenate((np.random.rand(30), [100.]))
with pytest.warns(UserWarning, match='likely bad'):
bar = ea.rt_chisq(foo_bad)
def test_rt_chisq():
"""Test reaction time chi-square fitting."""
# 1D should return single float
foo = np.random.rand(30)
assert_raises(ValueError, ea.rt_chisq, foo - 1.)
assert_equal(np.array(ea.rt_chisq(foo)).shape, ())
# 2D should return array with shape of input but without ``axis`` dimension
foo = np.random.rand(30).reshape((2, 3, 5))
for axis in range(-1, foo.ndim):
bar = ea.rt_chisq(foo, axis=axis)
assert_true(
np.all(np.equal(np.delete(foo.shape, axis), np.array(bar.shape))))
foo_bad = np.concatenate((np.random.rand(30), [100.]))
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter('always')
bar = ea.rt_chisq(foo_bad)
assert_equal(len(w), 1) # warn that there was a likely bad value
def test_rt_chisq():
"""Test reaction time chi-square fitting."""
# 1D should return single float
foo = np.random.rand(30)
assert_raises(ValueError, ea.rt_chisq, foo - 1.)
assert_equal(np.array(ea.rt_chisq(foo)).shape, ())
# 2D should return array with shape of input but without ``axis`` dimension
foo = np.random.rand(30).reshape((2, 3, 5))
for axis in range(-1, foo.ndim):
bar = ea.rt_chisq(foo, axis=axis)
assert_true(np.all(np.equal(np.delete(foo.shape, axis),
np.array(bar.shape))))
foo_bad = np.concatenate((np.random.rand(30), [100.]))
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter('always')
bar = ea.rt_chisq(foo_bad)
assert_equal(len(w), 1) # warn that there was a likely bad value
def chsq_rt(x):
# peak of chi squared distribution fit to data
y = np.array(x)
y = y[np.logical_not(np.isnan(y))]
if y.size == 0:
return np.nan
else:
return efa.rt_chisq(y)
#
corrects = list()
raw_rt = list()
for d in exp_data:
assert 'TONE_' in d['trial_id'][0][0] # prevent stupidity
# code hits 0, false alarms 1, misses 2, correct rejections 3
ti = ((d['trial_id'][0][0] == 'TONE_0') * 2 +
(len(d['keypress']) == 0))
corrects.append(True if ti in [0, 3] else False)
raw_rt.extend([float(d['keypress'][-1][1]) -
float(d['flip'][0][0])] if ti == 0 else [])
hmfx[si, ti] += 1
raw_rt = np.array(raw_rt)
raw_rt = raw_rt[raw_rt < 5.0] # cut off due to trial_ok not being recorded
assert len(raw_rt) > 20
avg_rt[si] = rt_chisq(raw_rt)
hitrate = hmfx[si, 0] / float(np.sum(hmfx[si, :2]))
farate = hmfx[si, 2] / float(np.sum(hmfx[si, 2:]))
print(' RT/HR/FAR: %s, %s, %s' % (avg_rt[si], hitrate, farate))
assert np.sum(corrects) == np.sum(hmfx[si, [0, 3]])
assert hmfx[si].sum() == 300
#
# read in pupillometry data for PRF
#
fnames = glob.glob(op.join(work_dir, 'data', subj + '_2014*', '*.edf'))
fnames.sort()
assert len(fnames) == 5 # screen luminance and 4x beeps
raws = list()
events = list()
for fname in fnames[1:]:
raw = pyeparse.RawEDF(fname)
#
corrects = list()
raw_rt = list()
for d in exp_data:
assert 'TONE_' in d['trial_id'][0][0] # prevent stupidity
# code hits 0, false alarms 1, misses 2, correct rejections 3
ti = ((d['trial_id'][0][0] == 'TONE_0') * 2
+ (len(d['keypress']) == 0))
corrects.append(True if ti in [0, 3] else False)
raw_rt.extend([float(d['keypress'][-1][1]) - float(d['flip'][0][0])
] if ti == 0 else [])
hmfx[si, ti] += 1
raw_rt = np.array(raw_rt)
raw_rt = raw_rt[raw_rt < 5.0] # cut off due to trial_ok not being recorded
assert len(raw_rt) > 20
avg_rt[si] = rt_chisq(raw_rt)
hitrate = hmfx[si, 0] / float(np.sum(hmfx[si, :2]))
farate = hmfx[si, 2] / float(np.sum(hmfx[si, 2:]))
print(' RT/HR/FAR: %s, %s, %s' % (avg_rt[si], hitrate, farate))
assert np.sum(corrects) == np.sum(hmfx[si, [0, 3]])
assert hmfx[si].sum() == 300
#
# read in pupillometry data for PRF
#
fnames = glob.glob(op.join(work_dir, 'data', subj + '_2014*', '*.edf'))
fnames.sort()
assert len(fnames) == 5 # screen luminance and 4x beeps
raws = list()
events = list()
for fname in fnames[1:]:
raw = pyeparse.RawEDF(fname)
