def test_tester_do_test_pvalue_prob(self):
t = mmct.tester()
t.test_statistic = 'Prob'
t.n_samples = 5
# null prob: [0.05,0.6,0.1,0.25]
# Sample 0: [0,4,2,2] Prob = 0.03402
# Sample 1: [1,5,2,0] Prob = 0.00653184
# Sample 2: [0,3,2,3] Prob = 0.0189
# Sample 3: [0,6,2,0] Prob = 0.01306368
# Sample 4: [1,1,5,1] Prob = 0.0000252
t.statistics = np.array(
[0.03402, 0.00653184, 0.0189, 0.01306368, 0.0000252])
t.fix = True
x = np.array([1, 3, 1, 3]) # Prob = 0.0189
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 0.8)
x = np.array([8, 0, 0, 0]) # Prob = 3.90625 × 10^-11
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 0.0)
x = np.array([1, 2, 2, 3]) # Prob = 0.004725
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 0.2)
def test_tester_do_test_pvalue_llr(self):
t = mmct.tester()
t.n_samples = 5
# null prob: [0.05,0.6,0.1,0.25]
# Sample 0: [0,4,2,2] LLR = -1.1032952365724916
# Sample 1: [1,5,2,0] LLR = -2.9529821682237408
# Sample 2: [0,3,2,3] LLR = -1.6389659003355966
# Sample 3: [0,6,2,0] LLR = -3.1714427716335686
# Sample 4: [1,1,5,1] LLR = -7.8174349521419151
t.statistics = np.array([
-1.1032952365724916, -2.9529821682237408, -1.6389659003355966,
-3.1714427716335686, -7.8174349521419151
])
t.fix = True
x = np.array([1, 3, 1, 3]) # LLR = -0.9458187197756513
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 1.0)
x = np.array([8, 0, 0, 0]) # LLR = -23.965858188431927
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 0.0)
x = np.array([1, 2, 2, 3]) # LLR = -2.2143300452391584
p = t.do_test(x, np.array([0.05, 0.6, 0.1, 0.25]))
self.assertEqual(p, 0.6)
def test_tester_do_test_params_llr(self):
t = mmct.tester()
t.n_samples = 200
t.statistics = np.zeros(80)
x = np.array([3, 4, 5, 6])
t.do_test(x, np.array([0.2, 0.25, 0.3, 0.25]))
self.assertEqual(t.statistics.size, 200)
def test_dice(rolls):
act_rolls = np.array(list(rolls.values()))
exp_freq = np.array([
1 / 36, 2 / 36, 3 / 36, 4 / 36, 5 / 36, 6 / 36, 5 / 36, 4 / 36, 3 / 36,
2 / 36, 1 / 36
])
t = mmct.tester()
t.n_trials = 10000
p = t.do_test(act_rolls, exp_freq)
print("H0: Counts = Expected counts")
print("Test: Monte Carlo two-sided multinomial test: p = %s" % p)
def test_tester_do_test_no_rerun_when_fixed(self):
t = mmct.tester()
t.n_samples = 4
x = np.array([3, 4, 5, 6])
t.do_test(x, np.array([0.2, 0.25, 0.3, 0.25]))
# Set artificial and impossible statistics. If generate_samples is run, these
# values will be overwritten, since they cannot occur mathematically
t.statistics = np.array([10, 12, 14, 16])
t.fix = True
y = np.array([6, 5, 4, 3])
t.do_test(y, np.array([0.2, 0.25, 0.3, 0.25]))
self.assertEqual(t.statistics.size, 4)
self.assertEqual(t.statistics[0], 10)
self.assertEqual(t.statistics[1], 12)
self.assertEqual(t.statistics[2], 14)
self.assertEqual(t.statistics[3], 16)
def test_tester_do_test_error_x_probs_not_same_dim(self):
t = mmct.tester()
x = np.array([3, 4, 5])
p = np.array([0.3, 0.6, 0.05, 0.05])
self.assertRaises(ValueError, t.do_test, x, p)
import numpy as np
import sys
sys.path.append("src")
# Hypothsised probability
# Cumulative: 0.10 0.17 0.48 0.52 0.63 0.87, 0.89 1.00
p = np.array([0.10, 0.07, 0.31, 0.04, 0.11, 0.24, 0.02, 0.11])
# Observations
x = np.array([17, 6, 30, 4, 8, 18, 1, 14])
t0 = time.time()
tx = mmct.tester()
tx.n_samples = 30000
p1 = tx.do_test(x, p)
print("Calculated p-value: {:.2f}".format(p1))
t1 = time.time()
ty = mmct.mt_tester()
ty.test_statistic = 'Prob'
ty.n_samples = 30000
p2 = ty.do_test(x, p)
print("Calculated p-value: {:.2f}".format(p2))
t2 = time.time()
