def run_on_batch(est, args):
s = sm.Simulation(args.sim_name)
pretty.print_namespace(s)
print()
print('batch=', args.batch_num)
print(est)
est.run_and_save_results(args.batch_num, s)
def run_on_batch(est, args):
sim = SumstatSimulation(args.sim_name)
pretty.print_namespace(sim)
print()
print("batch=", args.batch_num)
print(est)
est.run_and_save_results(args.batch_num, sim)
def main(args):
np.random.seed(args.beta_num + args.sample_num * 10000)
sim = SumstatSimulation(args.sim_name)
d = Dataset(sim.dataset)
pretty.print_namespace(sim)
print()
# read in noiseless phenotypes
Y = pickle.load(sim.noiseless_Y_file(args.beta_num))
# choose individuals and create ensemble of Ys
indices = np.random.choice(Y.shape[0], size=(sim.sample_size, ))
Y = Y[indices]
# compute how much noise to add
sigma2e = 1 - sim.h2g
print('adding noise. sigma2e =', sigma2e)
Y += np.sqrt(sigma2e) * np.random.randn(*Y.shape)
if sim.condition_on_covariates:
print('projecting covariates out of Y')
Y = d.project_out_covariates(Y, covariates=d.covariates[indices])
alphahat = np.zeros(d.M)
t0 = time()
def compute_sumstats_for_slice(s):
# X will be N x M
print(int(time() - t0), ': getting genotypes from file. SNPs', s)
X = d.get_standardized_genotypes(s)[indices]
if sim.condition_on_covariates:
print(int(time() - t0), ': projecting out covariates')
X = d.project_out_covariates(X, covariates=d.covariates[indices])
print(int(time() - t0), ': computing sumstats. SNPs', s)
alphahat[s[0]:s[1]] = X.T.dot(Y) / sim.sample_size
del X
map(compute_sumstats_for_slice, d.slices())
# write output
def write_output():
pickle.dump(indices,
sim.individuals_file(args.beta_num, args.sample_num, 'wb'),
2)
pickle.dump(Y, sim.noisy_Y_file(args.beta_num, args.sample_num, 'wb'),
2)
pickle.dump(alphahat,
sim.sumstats_file(args.beta_num, args.sample_num, 'wb'), 2)
write_output()
def main(args):
np.random.seed(args.beta_num + args.sample_num * 10000)
sim = SumstatSimulation(args.sim_name)
d = Dataset(sim.dataset)
pretty.print_namespace(sim); print()
# read in noiseless phenotypes
Y = pickle.load(sim.noiseless_Y_file(args.beta_num))
# choose individuals and create ensemble of Ys
indices = np.random.choice(Y.shape[0], size=(sim.sample_size,))
Y = Y[indices]
# compute how much noise to add
sigma2e = 1 - sim.h2g
print('adding noise. sigma2e =', sigma2e)
Y += np.sqrt(sigma2e) * np.random.randn(*Y.shape)
if sim.condition_on_covariates:
print('projecting covariates out of Y')
Y = d.project_out_covariates(Y, covariates=d.covariates[indices])
alphahat = np.zeros(d.M)
t0 = time()
def compute_sumstats_for_slice(s):
# X will be N x M
print(int(time() - t0), ': getting genotypes from file. SNPs', s)
X = d.get_standardized_genotypes(s)[indices]
if sim.condition_on_covariates:
print(int(time() - t0), ': projecting out covariates')
X = d.project_out_covariates(X, covariates=d.covariates[indices])
print(int(time() - t0), ': computing sumstats. SNPs', s)
alphahat[s[0]:s[1]] = X.T.dot(Y) / sim.sample_size
del X
map(compute_sumstats_for_slice, d.slices())
# write output
def write_output():
pickle.dump(indices, sim.individuals_file(
args.beta_num, args.sample_num, 'wb'), 2)
pickle.dump(Y, sim.noisy_Y_file(
args.beta_num, args.sample_num, 'wb'), 2)
pickle.dump(alphahat, sim.sumstats_file(
args.beta_num, args.sample_num, 'wb'), 2)
write_output()
def run_on_batch(est, args):
s = sm.Simulation(args.sim_name)
pretty.print_namespace(s); print()
print('batch=', args.batch_num)
print(est)
est.run_and_save_results(args.batch_num, s)
