def test_regression_gradient():
random = RandomState(94584)
N = 50
X = random.randn(N, 100)
offset = 0.5
mean = OffsetMean()
mean.offset = offset
mean.fix('offset')
mean.set_data(N)
cov = LinearCov()
cov.scale = 1.0
cov.set_data((X, X))
y = random.randn(N)
lmm = SlowLMM(y, mean, cov)
def func(x):
cov.scale = exp(x[0])
return lmm.value(mean.feed().value(), cov.feed().value())
def grad(x):
cov.scale = exp(x[0])
return lmm.gradient(mean.feed().value(),
cov.feed().value(),
mean.feed().gradient(),
cov.feed().gradient())
assert_almost_equal(check_grad(func, grad, [0]), 0)
def _mean_cov(offset, G, heritability, causal_variants, causal_variance,
random_state):
nsamples = G.shape[0]
G = stdnorm(G, axis=0)
G /= sqrt(G.shape[1])
mean1 = OffsetMean()
mean1.offset = offset
cov1 = LinearCov()
cov2 = EyeCov()
cov = SumCov([cov1, cov2])
mean1.set_data(nsamples, 'sample')
cov1.set_data((G, G), 'sample')
a = arange(nsamples)
cov2.set_data((a, a), 'sample')
cov1.scale = heritability - causal_variance
cov2.scale = 1 - heritability - causal_variance
means = [mean1]
if causal_variants is not None:
means += [_causal_mean(causal_variants, causal_variance, random_state)]
mean = SumMean(means)
return mean, cov
def test_fast_scan():
random = np.random.RandomState(9458)
N = 500
X = random.randn(N, N + 1)
X -= X.mean(0)
X /= X.std(0)
X /= np.sqrt(X.shape[1])
offset = 1.0
mean = OffsetMean()
mean.offset = offset
mean.set_data(N, purpose='sample')
cov_left = LinearCov()
cov_left.scale = 1.5
cov_left.set_data((X, X), purpose='sample')
cov_right = EyeCov()
cov_right.scale = 1.5
cov_right.set_data((arange(N), arange(N)), purpose='sample')
cov = SumCov([cov_left, cov_right])
lik = DeltaProdLik()
y = GLMMSampler(lik, mean, cov).sample(random)
(Q0, Q1), S0 = economic_qs_linear(X)
flmm = FastLMM(y, Q0, Q1, S0, covariates=ones((N, 1)))
flmm.learn(progress=False)
markers = random.randn(N, 2)
flmm_ = flmm.copy()
flmm_.M = concatenate([flmm.M, markers[:, 0][:, newaxis]], axis=1)
lml0 = flmm_.lml()
flmm_ = flmm.copy()
flmm_.M = concatenate([flmm.M, markers[:, 1][:, newaxis]], axis=1)
lml1 = flmm_.lml()
lik_trick = flmm.get_normal_likelihood_trick()
lmls = lik_trick.fast_scan(markers)[0]
assert_allclose(lmls, [lml0, lml1], rtol=1e-5)
def test_slowlmm_value_1():
random = RandomState(94584)
N = 50
X = random.randn(N, 100)
offset = 0.5
mean = OffsetMean()
mean.offset = offset
mean.fix('offset')
mean.set_data(N)
cov = LinearCov()
cov.scale = 1.0
cov.set_data((X, X))
y = random.randn(N)
lmm = SlowLMM(y, mean, cov)
assert_almost_equal(lmm.feed().value(), -153.623791551399108)
def test_maximize_2():
random = RandomState(94584)
N = 50
X = random.randn(N, 100)
offset = 0.5
mean = OffsetMean()
mean.offset = offset
mean.set_data(N)
cov = LinearCov()
cov.scale = 1.0
cov.set_data((X, X))
y = random.randn(N)
lmm = SlowLMM(y, mean, cov)
lmm.feed().maximize()
assert_almost_equal(lmm.feed().value(), -79.365136339619610)
def test_GLMMSampler_binomial():
random = RandomState(4503)
X = random.randn(10, 15)
link = LogitLink()
lik = BinomialLik(5, link)
mean = OffsetMean()
mean.offset = 1.2
mean.set_data(10, 'sample')
cov = LinearCov()
cov.set_data((X, X), 'sample')
sampler = GLMMSampler(lik, mean, cov)
assert_equal(sampler.sample(random), [0, 5, 0, 5, 1, 1, 5, 0, 5, 5])
mean.offset = 0.
assert_equal(sampler.sample(random), [5, 4, 1, 0, 0, 1, 4, 5, 5, 0])
mean = OffsetMean()
mean.offset = 0.0
mean.set_data(10, 'sample')
cov1 = LinearCov()
cov1.set_data((X, X), 'sample')
cov2 = EyeCov()
a = arange(10)
cov2.set_data((a, a), 'sample')
cov1.scale = 1e-4
cov2.scale = 1e-4
cov = SumCov([cov1, cov2])
lik = BinomialLik(100, link)
sampler = GLMMSampler(lik, mean, cov)
assert_equal(
sampler.sample(random), [56, 56, 55, 51, 59, 45, 47, 43, 51, 38])
cov2.scale = 100.
sampler = GLMMSampler(lik, mean, cov)
assert_equal(
sampler.sample(random), [99, 93, 99, 75, 77, 0, 0, 100, 99, 12])
def test_GLMMSampler_poisson():
random = RandomState(4503)
X = random.randn(10, 15)
link = LogLink()
lik = PoissonLik(link)
mean = OffsetMean()
mean.offset = 1.2
mean.set_data(10, 'sample')
cov = LinearCov()
cov.set_data((X, X), 'sample')
sampler = GLMMSampler(lik, mean, cov)
assert_equal(
sampler.sample(random), [0, 289, 0, 11, 0, 0, 176, 0, 228, 82])
mean = OffsetMean()
mean.offset = 0.0
mean.set_data(10, 'sample')
cov1 = LinearCov()
cov1.set_data((X, X), 'sample')
cov2 = EyeCov()
a = arange(10)
cov2.set_data((a, a), 'sample')
cov1.scale = 1e-4
cov2.scale = 1e-4
cov = SumCov([cov1, cov2])
sampler = GLMMSampler(lik, mean, cov)
assert_equal(sampler.sample(random), [2, 0, 1, 2, 1, 1, 1, 2, 0, 0])
cov2.scale = 100.
sampler = GLMMSampler(lik, mean, cov)
assert_equal(sampler.sample(random), [0, 0, 0, 0, 1, 0, 0, 1196, 0, 0])
def test_learn():
random = np.random.RandomState(9458)
N = 500
X = random.randn(N, N + 1)
X -= X.mean(0)
X /= X.std(0)
X /= np.sqrt(X.shape[1])
offset = 1.0
mean = OffsetMean()
mean.offset = offset
mean.set_data(N, purpose='sample')
cov_left = LinearCov()
cov_left.scale = 1.5
cov_left.set_data((X, X), purpose='sample')
cov_right = EyeCov()
cov_right.scale = 1.5
cov_right.set_data((arange(N), arange(N)), purpose='sample')
cov = SumCov([cov_left, cov_right])
lik = DeltaProdLik()
y = GLMMSampler(lik, mean, cov).sample(random)
(Q0, Q1), S0 = economic_qs_linear(X)
flmm = FastLMM(y, Q0, Q1, S0, covariates=ones((N, 1)))
flmm.learn(progress=False)
assert_allclose(flmm.beta[0], 0.8997652129631661, rtol=1e-5)
assert_allclose(flmm.genetic_variance, 1.7303981309775553, rtol=1e-5)
assert_allclose(flmm.environmental_variance, 1.2950028351268132, rtol=1e-5)
def test_maximize_1():
random = RandomState(94584)
N = 50
X = random.randn(N, 100)
offset = 0.5
mean = OffsetMean()
mean.offset = offset
mean.fix('offset')
mean.set_data(N)
cov = LinearCov()
cov.scale = 1.0
cov.set_data((X, X))
y = random.randn(N)
lmm = SlowLMM(y, mean, cov)
m = mean.feed().value()
K = cov.feed().value()
assert_almost_equal(lmm.value(m, K), -153.62379155139911)
lmm.feed().maximize()
assert_almost_equal(lmm.feed().value(), -79.899212241487518)