def test_multinomial_elementwise_distribution(self):
'''Verify that the created variables approach a multinomial distribution for large numbers
of samples.'''
(m, n, k) = (6, 5, 1)
r = 2 ** np.arange(4, 17)
p = statutil.random_row_stochastic((m, n))
#p = statutil.scale_row_sums(np.ones((m, n)))
error = np.zeros((len(r),))
for (i, r_val) in enumerate(r):
for _ in xrange(k):
x = statutil.multinomial_elementwise(p, r_val)
# Root-mean-square-error of observed frequencies w.r.t. desired frequencies
error[i] += statutil.norm_frobenius_scaled(statutil.hist(x, n) / (1.0 * r_val) - p)
error[i] /= (1.0 * k)
# Validate the model error of the central limit theorem: C*r^(-0.5).
# This is a consequence of the Central Limit Theorem. We are making k experiments for
# each value of n. Even if k=1, there's a 95% chance that we are within ~1.6 standard deviations
# from the mean of the normal distribution sqrt(n)*[observed freq variable - p[i,j]] for each
# entry j of a row i of the matrix p. So if row i's stddev is s[i], the sum of squared errors
# should be (with 95% confidence) <= n * (1.96*s[i])^2. So
# C <= sqrt(sum(n * (1.5*s[i])^2)_i / (m*n)) = 1.96 * sqrt(s[i]^2/m).
# See http://en.wikipedia.org/wiki/Central_limit_theorem
alpha, c, r_value, _, _ = linregress(np.log(r), np.log(error))
c = np.exp(c)
# print c , 1.96 * np.linalg.linalg.norm(np.sum(p * np.arange(p.shape[1]) ** 2, axis=1) -
# np.sum(p * np.arange(p.shape[1]), axis=1) ** 2,
# 2) / np.sqrt(p.shape[0]),
assert_almost_equal(alpha, -0.5, decimal=1, err_msg='Unexpected error term growth power')
self.assertTrue(c <= 1.96 * np.linalg.linalg.norm(np.sum(p * np.arange(p.shape[1]) ** 2, axis=1) -
np.sum(p * np.arange(p.shape[1]), axis=1) ** 2,
2) / np.sqrt(p.shape[0]),
'Error term coefficient outside 95% confidence interval')
self.assertTrue(abs(r_value) > 0.99, 'Error does not fit a power law in sample size')
def test_multinomial_elementwise_vector(self):
'''Test creating multinomial variables (r=1).'''
(m, n) = (20, 5)
p = statutil.random_row_stochastic((m, n))
x = statutil.multinomial_elementwise(p)
assert_equal(x, [3, 1, 3, 4, 0, 2, 2, 3, 1, 4, 2, 3, 2, 1, 4, 2, 4, 3, 4, 3],
'Wrong random multinomial generated')
def test_multinomial_elementwise_vector(self):
'''Test creating multinomial variables (r=1).'''
(m, n) = (20, 5)
p = statutil.random_row_stochastic((m, n))
x = statutil.multinomial_elementwise(p)
assert_equal(
x, [3, 1, 3, 4, 0, 2, 2, 3, 1, 4, 2, 3, 2, 1, 4, 2, 4, 3, 4, 3],
'Wrong random multinomial generated')
def test_multinomial_elementwise_matrix(self):
'''Test creating multinomial variables (r > 1).'''
(m, n) = (20, 5)
p = statutil.random_row_stochastic((m, n))
x = statutil.multinomial_elementwise(p, 2)
assert_equal(x, [[3, 3], [1, 2], [3, 3], [4, 3], [0, 1], [2, 2], [2, 0], [3, 1], [1, 2],
[4, 0], [2, 2], [3, 1], [2, 1], [1, 1], [4, 2], [2, 3], [4, 3], [3, 4],
[4, 3], [3, 1]],
'Wrong random multinomial generated')
def _simulate_imputation(self):
'''Simulate imputing i from phased T-haplotypes.'''
x = IMPUTABLE[statutil.multinomial_elementwise(self.p)]
a0 = np.any(x / 2) # Bit 0 set for at least one element of x ==> allele 0 is imputable
a1 = np.any(x % 2) # Bit 1 set for at least one element of x ==> allele 1 is imputable
if a0 & a1: # Both alleles are imputable
self.count_full += 1
if a0 | a1: # Either allele is imputable
self.count_part += 1
self.count += 1
def test_multinomial_elementwise_matrix(self):
'''Test creating multinomial variables (r > 1).'''
(m, n) = (20, 5)
p = statutil.random_row_stochastic((m, n))
x = statutil.multinomial_elementwise(p, 2)
assert_equal(x,
[[3, 3], [1, 2], [3, 3], [4, 3], [0, 1], [2, 2], [2, 0],
[3, 1], [1, 2], [4, 0], [2, 2], [3, 1], [2, 1], [1, 1],
[4, 2], [2, 3], [4, 3], [3, 4], [4, 3], [3, 1]],
'Wrong random multinomial generated')
def _simulate_imputation(self):
'''Simulate imputing i from phased T-haplotypes.'''
x = IMPUTABLE[statutil.multinomial_elementwise(self.p)]
a0 = np.any(
x / 2
) # Bit 0 set for at least one element of x ==> allele 0 is imputable
a1 = np.any(
x % 2
) # Bit 1 set for at least one element of x ==> allele 1 is imputable
if a0 & a1: # Both alleles are imputable
self.count_full += 1
if a0 | a1: # Either allele is imputable
self.count_part += 1
self.count += 1
def test_multinomial_elementwise_distribution(self):
'''Verify that the created variables approach a multinomial distribution for large numbers
of samples.'''
(m, n, k) = (6, 5, 1)
r = 2**np.arange(4, 17)
p = statutil.random_row_stochastic((m, n))
#p = statutil.scale_row_sums(np.ones((m, n)))
error = np.zeros((len(r), ))
for (i, r_val) in enumerate(r):
for _ in xrange(k):
x = statutil.multinomial_elementwise(p, r_val)
# Root-mean-square-error of observed frequencies w.r.t. desired frequencies
error[i] += statutil.norm_frobenius_scaled(
statutil.hist(x, n) / (1.0 * r_val) - p)
error[i] /= (1.0 * k)
# Validate the model error of the central limit theorem: C*r^(-0.5).
# This is a consequence of the Central Limit Theorem. We are making k experiments for
# each value of n. Even if k=1, there's a 95% chance that we are within ~1.6 standard deviations
# from the mean of the normal distribution sqrt(n)*[observed freq variable - p[i,j]] for each
# entry j of a row i of the matrix p. So if row i's stddev is s[i], the sum of squared errors
# should be (with 95% confidence) <= n * (1.96*s[i])^2. So
# C <= sqrt(sum(n * (1.5*s[i])^2)_i / (m*n)) = 1.96 * sqrt(s[i]^2/m).
# See http://en.wikipedia.org/wiki/Central_limit_theorem
alpha, c, r_value, _, _ = linregress(np.log(r), np.log(error))
c = np.exp(c)
# print c , 1.96 * np.linalg.linalg.norm(np.sum(p * np.arange(p.shape[1]) ** 2, axis=1) -
# np.sum(p * np.arange(p.shape[1]), axis=1) ** 2,
# 2) / np.sqrt(p.shape[0]),
assert_almost_equal(alpha,
-0.5,
decimal=1,
err_msg='Unexpected error term growth power')
self.assertTrue(
c <= 1.96 * np.linalg.linalg.norm(
np.sum(p * np.arange(p.shape[1])**2, axis=1) -
np.sum(p * np.arange(p.shape[1]), axis=1)**2, 2) /
np.sqrt(p.shape[0]),
'Error term coefficient outside 95% confidence interval')
self.assertTrue(
abs(r_value) > 0.99,
'Error does not fit a power law in sample size')
def __handle_fill_missing_genotypes(self, request):
'''Fill missing genotype entries by randomly sampling from the multinomial distribution with
estimated genotype frequencies at the corresponding SNP.'''
# Load problem fields
if request.params.debug:
print 'Filling missing genotypes from estimated genotype distribution'
problem = request.problem
g = problem.genotype.data
snp_frequency = problem.info.snp['frequency'][:, FILLED_GENOTYPES]
# Recode genotypes to a single number
r = recode.recode_single_genotype(g)
# Find SNP, sample indices of missing data
missing = recode.where_missing(r)
# Generate random multinomial values; map them to genotype codes
filled_code = multinomial_elementwise(scale_row_sums(snp_frequency[missing[SNP]])) + 2
# Fill-in all genotypes of a certain value in a vectorized manner
for (genotype, code) in recode.GENOTYPE_CODE.iteritems():
index = np.where(filled_code == code)[0]
g[missing[SNP][index], missing[SAMPLE][index], :] = genotype
return False
def __handle_fill_missing_genotypes(self, request):
'''Fill missing genotype entries by randomly sampling from the multinomial distribution with
estimated genotype frequencies at the corresponding SNP.'''
# Load problem fields
if request.params.debug:
print 'Filling missing genotypes from estimated genotype distribution'
problem = request.problem
g = problem.genotype.data
snp_frequency = problem.info.snp['frequency'][:, FILLED_GENOTYPES]
# Recode genotypes to a single number
r = recode.recode_single_genotype(g)
# Find SNP, sample indices of missing data
missing = recode.where_missing(r)
# Generate random multinomial values; map them to genotype codes
filled_code = multinomial_elementwise(
scale_row_sums(snp_frequency[missing[SNP]])) + 2
# Fill-in all genotypes of a certain value in a vectorized manner
for (genotype, code) in recode.GENOTYPE_CODE.iteritems():
index = np.where(filled_code == code)[0]
g[missing[SNP][index], missing[SAMPLE][index], :] = genotype
return False
