def test_multiple_comparisons(self):
"""multiple_comparisons should match values from R"""
self.assertFloatEqual(multiple_comparisons(1e-7, 10000), 1 - 0.9990005)
self.assertFloatEqual(multiple_comparisons(0.05, 10), 0.4012631)
self.assertFloatEqual(multiple_comparisons(1e-20, 1), 1e-20)
self.assertFloatEqual(multiple_comparisons(1e-300, 1), 1e-300)
self.assertFloatEqual(multiple_comparisons(0.95, 3), 0.99987499999999996)
self.assertFloatEqual(multiple_comparisons(0.75, 100), 0.999999999999679)
self.assertFloatEqual(multiple_comparisons(0.5, 1000), 1)
self.assertFloatEqual(multiple_comparisons(0.01, 1000), 0.99995682875259)
self.assertFloatEqual(multiple_comparisons(0.5, 5), 0.96875)
self.assertFloatEqual(multiple_comparisons(1e-20, 10), 1e-19)
def getPValueLoose(self, module, alphabet_len=None):
"""Returns the Pvalue of a module.
- pass alphabet_len if alphabet other than module.Alphabet was used
to find modules (i.e. using a protein reduced alphabet)
"""
# Length of the module
module_len = len(module.Template)
# Get alphabet
alphabet = module.Alphabet
# if alphabet length has not been passed, get it from module.Alphabet
#
if not alphabet_len:
alphabet_len = len(alphabet.Order)
# Total length of the alignment
aln_len = sum(map(len, self.Alignment.values()))
# Number of sequences in the alignment
num_seqs = len(self.Alignment)
# get module_p
module_p = 1
for char in module.Template:
module_p *= self._degen_p(char, module.Alphabet)
# Mean for passing to poisson_high using loose correction
loose_mean = (aln_len + num_seqs * (1 - module_len)) * (module_p)
# Loose P value from poisson_high
loose_p_value = poisson_high(len(module) - 1, loose_mean)
# Correct P value for multiple comparisons
loose_p_corrected = multiple_comparisons(loose_p_value, alphabet_len ** module_len)
return loose_p_corrected
