def boot_pliv(theta, Y, D, Z, g_hat, m_hat, r_hat, smpls, score, se, bootstrap, n_rep, dml_procedure):
n_obs = len(Y)
weights = draw_weights(bootstrap, n_rep, n_obs)
if np.isscalar(theta):
n_d = 1
else:
n_d = len(theta)
assert n_d == 1
boot_theta, boot_t_stat = boot_pliv_single_treat(theta, Y, D, Z, g_hat, m_hat, r_hat,
smpls, score, se, weights, n_rep, dml_procedure)
return boot_theta, boot_t_stat
def boot_plr(theta,
Y,
D,
g_hat,
m_hat,
smpls,
score,
se,
bootstrap,
n_rep,
dml_procedure,
apply_cross_fitting=True):
if apply_cross_fitting:
n_obs = len(Y)
else:
test_index = smpls[0][1]
n_obs = len(test_index)
weights = draw_weights(bootstrap, n_rep, n_obs)
if np.isscalar(theta):
n_d = 1
else:
n_d = len(theta)
if n_d > 1:
boot_theta = np.full((n_d, n_rep), np.nan)
boot_t_stat = np.full((n_d, n_rep), np.nan)
for i_d in range(n_d):
boot_theta[i_d, :], boot_t_stat[i_d, :] = boot_plr_single_treat(
theta[i_d], Y, D[:, i_d], g_hat[i_d], m_hat[i_d], smpls, score,
se[i_d], weights, n_rep, dml_procedure, apply_cross_fitting)
else:
boot_theta, boot_t_stat = boot_plr_single_treat(
theta, Y, D, g_hat, m_hat, smpls, score, se, weights, n_rep,
dml_procedure, apply_cross_fitting)
return boot_theta, boot_t_stat