def correlation_p_pearsonr2(data0, data1):
r = stats.pearsonr(data0, data1)
if math.isnan(r):
return float('NaN')
if r == 1.:
return 0.
n = len(data0)
assert n == len(data1)
# Compute observed t statistic.
t = r * math.sqrt((n - 2)/(1 - r**2))
# Compute p-value for two-sided t-test.
return 2 * stats.t_cdf(-abs(t), n - 2)
def correlation_p_pearsonr2(data0, data1):
r = stats.pearsonr(data0, data1)
if math.isnan(r):
return float('NaN')
if r == 1.:
return 0.
n = len(data0)
assert n == len(data1)
# Compute observed t statistic.
t = r * math.sqrt((n - 2) / (1 - r**2))
# Compute p-value for two-sided t-test.
return 2 * stats.t_cdf(-abs(t), n - 2)
def test_t_cdf():
# Non-positive degrees of freedom should throw an error.
with pytest.raises(ValueError):
stats.t_cdf(0,0)
with pytest.raises(ValueError):
stats.t_cdf(2,-10)
# CDF of x = 0 should be 0.5.
assert relerr(.5, stats.t_cdf(0,12)) < .01
assert relerr(.5, stats.t_cdf(0,6)) < .01
assert relerr(.5, stats.t_cdf(0,130)) < .01
# Test against various reference values.
assert relerr(.57484842931039226, stats.t_cdf(.8, .1)) < .05
assert relerr(.64922051214061649, stats.t_cdf(.6, .6)) < .05
assert relerr(.51046281131211058, stats.t_cdf(.1, .05)) < .05
assert relerr(.99999944795492968, stats.t_cdf(9, 12)) < .05
assert relerr(.92318422834700042, stats.t_cdf(1.9, 3)) < .05
assert relerr(.81430689864299455, stats.t_cdf(1, 4.2)) < .05
assert relerr(.99995442539414559, stats.t_cdf(8, 7)) < .05
assert relerr(.92010336338282994, stats.t_cdf(3.9, 1)) < .05
assert relerr(1.0, stats.t_cdf(193, 121)) < .05
assert relerr(.42515157068960779, stats.t_cdf(-.8, .1)) < .05
assert relerr(.35077948785938345, stats.t_cdf(-.6, .6)) < .05
assert relerr(.48953718868788948, stats.t_cdf(-.1, .05)) < .05
assert relerr(.076815771652999562, stats.t_cdf(-1.9, 3)) < .05
assert relerr(.18569310135700545, stats.t_cdf(-1, 4.2)) < .05
assert relerr(.17530833141010374, stats.t_cdf(-1, 7)) < .05
assert relerr(.079896636617170003, stats.t_cdf(-3.9, 1)) < .05
assert relerr(.30899158341328747, stats.t_cdf(-0.5, 121)) < .05
# Test against reference very close to zero.
# XXX Why are we testing chi2_sf here?
assert relerr(.346437e-4, stats.chi2_sf(193,121)) < .01