def test_sem_and_variability():
"Test variability() and standard_error_of_the_mean() functions"
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x
x = ds['exposure'].as_factor()
match = ds['subject']
# invalid spec
assert_raises(ValueError, stats.variability, y, 0, 0, '1mile', 0)
assert_raises(ValueError, stats.variability, y, 0, 0, 'ci7ci', 0)
# standard error
target = scipy.stats.sem(y, 0, 1)
e = stats.variability(y, None, None, 'sem', False)
assert_almost_equal(e, target)
e = stats.variability(y, None, None, '2sem', False)
assert_almost_equal(e, 2 * target)
# within subject standard-error
target = scipy.stats.sem(stats.residuals(y[:, None], match), 0, len(match.cells))
assert_almost_equal(stats.variability(y, None, match, 'sem', True), target)
assert_almost_equal(stats.variability(y, None, match, 'sem', False), target)
# one data point per match cell
n = match.df + 1
assert_raises(ValueError, stats.variability, y[:n], None, match[:n], 'sem', True)
target = np.array([scipy.stats.sem(y[x == cell], 0, 1) for cell in x.cells])
es = stats.variability(y, x, None, 'sem', False)
assert_allclose(es, target)
stats.variability(y, x, None, 'sem', True)
# confidence intervals
stats.variability(y, None, None, '95%ci', False)
stats.variability(y, x, None, '95%ci', True)
stats.variability(y, x, match, '95%ci', True)
def test_confidence_interval():
"Test confidence_interval()"
from rpy2.robjects import r
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x[:, None]
x = ds['exposure'].as_factor()
subject = ds['subject']
ds.to_r('ds')
# simple confidence interval of the mean
ci = stats.confidence_interval(y)[0]
r("s <- sd(ds$n_recalled)")
r("n <- length(ds$n_recalled)")
ci_r = r("qt(0.975, df=n-1) * s / sqrt(n)")[0]
eq_(ci, ci_r)
# pooled variance
ci = stats.confidence_interval(y, x)[0]
assert_almost_equal(ci, 3.85, delta=0.01)
assert_raises(NotImplementedError, stats.confidence_interval, y[1:], x[1:])
# within subject confidence interval
ci = stats.confidence_interval(y, x, subject)[0]
assert_almost_equal(ci, 0.52, 2)
def test_confidence_interval():
"Test confidence_interval()"
from rpy2.robjects import r
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x[:, None].astype(np.float64)
x = ds['exposure'].as_factor()
subject = ds['subject']
ds.to_r('ds')
# simple confidence interval of the mean
ci = stats.confidence_interval(y)[0]
r("s <- sd(ds$n_recalled)")
r("n <- length(ds$n_recalled)")
ci_r = r("qt(0.975, df=n-1) * s / sqrt(n)")[0]
eq_(ci, ci_r)
# pooled variance
ci = stats.confidence_interval(y, x)[0]
assert_almost_equal(ci, 3.85, delta=0.01)
assert_raises(NotImplementedError, stats.confidence_interval, y[1:], x[1:])
# within subject confidence interval
ci = stats.confidence_interval(y, x, subject)[0]
assert_almost_equal(ci, 0.52, 2)
def test_timeplot():
"Test plot.Timeplot"
ds = datasets.get_loftus_masson_1994()
ds['cat'] = Factor([int(s) > 5 for s in ds['subject']], labels={True: 'a', False: 'b'})
plot.Timeplot('n_recalled', 'subject', 'exposure', ds=ds)
plot.Timeplot('n_recalled', 'cat', 'exposure', ds=ds)
plot.Timeplot('n_recalled', 'cat', 'exposure', 'subject', ds=ds, x_jitter=True)
def test_variability():
"Test variability functions"
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x.astype(np.float64)
x = ds['exposure'].as_factor()
match = ds['subject']
sem = scipy.stats.sem(y, 0, 1)
ci = sem * scipy.stats.t.isf(0.05 / 2., len(y) - 1)
# invalid spec
with pytest.raises(ValueError):
stats.variability(y, 0, 0, '1mile', 0)
with pytest.raises(ValueError):
stats.variability(y, 0, 0, 'ci7ci', 0)
# standard error
assert stats.variability(y, None, None, 'sem', False) == sem
assert stats.variability(y, None, None, '2sem', False) == 2 * sem
# within subject standard-error
target = scipy.stats.sem(stats.residuals(y[:, None], match), 0,
len(match.cells))[0]
assert stats.variability(y, None, match, 'sem',
True) == pytest.approx(target)
assert stats.variability(y, None, match, 'sem',
False) == pytest.approx(target)
# one data point per match cell
n = match.df + 1
with pytest.raises(ValueError):
stats.variability(y[:n], None, match[:n], 'sem', True)
target = np.array(
[scipy.stats.sem(y[x == cell], 0, 1) for cell in x.cells])
es = stats.variability(y, x, None, 'sem', False)
assert_allclose(es, target)
stats.variability(y, x, None, 'sem', True)
# confidence intervals
assert stats.variability(y, None, None, '95%ci',
False) == pytest.approx(ci)
assert stats.variability(y, x, None, '95%ci',
True) == pytest.approx(3.86,
abs=1e-2) # L&M: 3.85
assert stats.variability(y, x, match, '95%ci',
True) == pytest.approx(0.52, abs=1e-2)
assert_equal(
stats.variability(y, x, None, '95%ci', False)[::-1],
stats.variability(y, x, None, '95%ci', False, x.cells[::-1]))
def test_timeplot():
"Test plot.Timeplot"
ds = datasets.get_loftus_masson_1994()
ds['cat'] = Factor([int(s) > 5 for s in ds['subject']],
labels={
True: 'a',
False: 'b'
})
plot.Timeplot('n_recalled', 'subject', 'exposure', ds=ds)
plot.Timeplot('n_recalled', 'cat', 'exposure', ds=ds)
plot.Timeplot('n_recalled',
'cat',
'exposure',
'subject',
ds=ds,
x_jitter=True)
def test_sem_and_variability():
"Test variability() and standard_error_of_the_mean() functions"
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x.astype(np.float64)
x = ds['exposure'].as_factor()
match = ds['subject']
# invalid spec
assert_raises(ValueError, stats.variability, y, 0, 0, '1mile', 0)
assert_raises(ValueError, stats.variability, y, 0, 0, 'ci7ci', 0)
# standard error
target = scipy.stats.sem(y, 0, 1)
e = stats.variability(y, None, None, 'sem', False)
assert_almost_equal(e, target)
e = stats.variability(y, None, None, '2sem', False)
assert_almost_equal(e, 2 * target)
# within subject standard-error
target = scipy.stats.sem(stats.residuals(y[:, None], match), 0,
len(match.cells))
assert_almost_equal(stats.variability(y, None, match, 'sem', True), target)
assert_almost_equal(stats.variability(y, None, match, 'sem', False),
target)
# one data point per match cell
n = match.df + 1
assert_raises(ValueError, stats.variability, y[:n], None, match[:n], 'sem',
True)
target = np.array(
[scipy.stats.sem(y[x == cell], 0, 1) for cell in x.cells])
es = stats.variability(y, x, None, 'sem', False)
assert_allclose(es, target)
stats.variability(y, x, None, 'sem', True)
# confidence intervals
stats.variability(y, None, None, '95%ci', False)
stats.variability(y, x, None, '95%ci', True)
stats.variability(y, x, match, '95%ci', True)
assert_equal(
stats.variability(y, x, None, '95%ci', False)[::-1],
stats.variability(y, x, None, '95%ci', False, x.cells[::-1]))
def test_variability():
"Test variability functions"
ds = datasets.get_loftus_masson_1994()
y = ds['n_recalled'].x.astype(np.float64)
x = ds['exposure'].as_factor()
match = ds['subject']
sem = scipy.stats.sem(y, 0, 1)
ci = sem * scipy.stats.t.isf(0.05 / 2., len(y) - 1)
# invalid spec
assert_raises(ValueError, stats.variability, y, 0, 0, '1mile', 0)
assert_raises(ValueError, stats.variability, y, 0, 0, 'ci7ci', 0)
# standard error
assert_almost_equal(stats.variability(y, None, None, 'sem', False), sem)
assert_almost_equal(stats.variability(y, None, None, '2sem', False), 2 * sem)
# within subject standard-error
target = scipy.stats.sem(stats.residuals(y[:, None], match), 0, len(match.cells))
assert_almost_equal(stats.variability(y, None, match, 'sem', True), target)
assert_almost_equal(stats.variability(y, None, match, 'sem', False), target)
# one data point per match cell
n = match.df + 1
assert_raises(ValueError, stats.variability, y[:n], None, match[:n], 'sem', True)
target = np.array([scipy.stats.sem(y[x == cell], 0, 1) for cell in x.cells])
es = stats.variability(y, x, None, 'sem', False)
assert_allclose(es, target)
stats.variability(y, x, None, 'sem', True)
# confidence intervals
assert_almost_equal(stats.variability(y, None, None, '95%ci', False), ci)
assert_almost_equal(stats.variability(y, x, None, '95%ci', True), 3.86, 2) # L&M: 3.85
assert_almost_equal(stats.variability(y, x, match, '95%ci', True), 0.52, 2)
assert_equal(stats.variability(y, x, None, '95%ci', False)[::-1],
stats.variability(y, x, None, '95%ci', False, x.cells[::-1]))