def bad_alpha_sampler_simple_sim(n=[33, 21, 22, 22, 24, 11]):
# generate data and model
d = data.simple_hierarchical_data(n)
m = models.simple_hierarchical_model(d['y'])
# fit model with MCMC, but with badly initialized step method
mcmc = mc.MCMC(m)
mcmc.use_step_method(mc.AdaptiveMetropolis, [m['inv_sigma_sq'], m['mu'], m['inv_tau_sq']])
mcmc.use_step_method(mc.NoStepper, m['alpha'])
mcmc.sample(60000, 10000, 10)
return d, m
def good_simple_sim(n=[33, 21, 22, 22, 24, 11]):
# generate data and model
d = data.simple_hierarchical_data(n)
m = models.simple_hierarchical_model(d['y'])
# fit model with MCMC
mcmc = mc.MCMC(m)
mcmc.use_step_method(mc.AdaptiveMetropolis, [m['inv_sigma_sq'], m['mu'], m['inv_tau_sq']])
mcmc.use_step_method(mc.AdaptiveMetropolis, [m['alpha']])
mcmc.sample(60000, 10000, 10)
return d, m
def bad_mu_prior_simple_sim(n=[33, 21, 22, 22, 24, 11]):
# generate data and model, intentionally misspecifying prior on mu
d = data.simple_hierarchical_data(n)
m = models.simple_hierarchical_model(d['y'])
m['mu'].parents['mu'] = -5.
m['mu'].parents['tau'] = .001**-2
# fit model with MCMC
mcmc = mc.MCMC(m)
mcmc.use_step_method(mc.AdaptiveMetropolis, [m['inv_sigma_sq'], m['mu'], m['inv_tau_sq']])
mcmc.use_step_method(mc.AdaptiveMetropolis, [m['alpha']])
mcmc.sample(60000, 10000, 10)
return d, m
def test_simple_model(self):
p = [[1,2,3], [4,5,6,7]]
vars = models.simple_hierarchical_model(p)
assert 'y' in vars
def test_simple_model(self):
p = [[1, 2, 3], [4, 5, 6, 7]]
vars = models.simple_hierarchical_model(p)
assert 'y' in vars