def setup_class(cls):
endog_bin = (endog > endog.mean()).astype(int)
cls.cov_type = 'cluster'
mod1 = GLM(endog_bin, exog, family=families.Gaussian(link=links.CDFLink()))
cls.res1 = mod1.fit(cov_type='cluster', cov_kwds=dict(groups=group))
mod1 = smd.Probit(endog_bin, exog)
cls.res2 = mod1.fit(cov_type='cluster', cov_kwds=dict(groups=group))
def setup_class(cls):
endog_bin = (endog > endog.mean()).astype(int)
cls.cov_type = 'cluster'
mod1 = GLM(endog_bin, exog, family=families.Binomial(link=links.probit()))
cls.res1 = mod1.fit(method='newton',
cov_type='cluster', cov_kwds=dict(groups=group))
mod1 = smd.Probit(endog_bin, exog)
cls.res2 = mod1.fit(cov_type='cluster', cov_kwds=dict(groups=group))
cls.rtol = 1e-6
def compute_marg_likelihood_and_NSE(X, y, iters, init, hypers):
''' Compute the marginal likelihood from the Gibbs Sampler output according to Chib (1995)
X (ndarray): exogeneous variables
y (array-like): endogeneous variables
iters (int): length of the MCMC
init (dict): initialisation parameters
hypers (array-like): hyper-parameters
returns (float): the marginal likelihood/normalizing constant
'''
# Initialisation
a, A = init['a'], init['A']
beta, beta_z, B = GibbsSampler(X, y, iters, init, hypers)
beta_star = np.array(beta).mean(axis=0)
beta_z = np.array(beta_z)
## Marginal likelihood computation P7, right column
# First term:
log_like = sm.Probit(endog=y, exog=X).loglike(params=beta_star)
# Second term
prior = multivariate_normal.logpdf(x=beta_star, mean=a, cov=A)
# Third term
conditional_densities = np.array([
multivariate_normal.pdf(x=beta_star, mean=beta_z[i], cov=B)
for i in range(iters)
])
posterior = np.log(conditional_densities.mean())
# pdf renvoie un gros nombre...: Compréhension de liste peut être amélioré
# Marginal likelihood
log_marg_likelihood = log_like + prior - posterior
#Numerical Standard Error
NSE = np.sqrt(
compute_var_h(conditional_densities, q=10) /
(conditional_densities.mean()**2))
return log_marg_likelihood, NSE
