Skip to content

Instantly share code, notes, and snippets.

@recluze

recluze/lin-reg-ppc.py

Created October 3, 2015 15:22
Show Gist options
  • Save recluze/5051735efe3fc189b90d to your computer and use it in GitHub Desktop.
Save recluze/5051735efe3fc189b90d to your computer and use it in GitHub Desktop.
Download ZIP
Raw
lin-reg-ppc.py
model = Model()
with model:
beta = Uniform('beta', lower=-10, upper=10, shape=(1, n_predictors))
p = tinvlogit(sum(beta * predictors, 1))
Y_obs = Bernoulli('Y_obs', p=p, observed=Y)
start = find_MAP()
step = NUTS(scaling=start)
trace = sample(mcmc_training_sample_size, step, start)
ppc_sample_size = 500
beta_rnd = trace[(-1 * ppc_sample_size):]['beta']
# get point to test and calculate posterior distribution
pred_test = get_point_to_test()
ps_np = tinvlogit(sum(beta_rnd * pred_test), 1)
posterior_mean = ps_np.mean()
posterior_sigma = ps_np.std()
posterior_interval = hpd(ps_np)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment