def test_get_df(self, df):
d = Data(
df=df,
col_t='time',
col_obs='ln_death_rate',
col_covs=['covariate_1'],
col_group='group',
obs_space=ln_gaussian_pdf,
)
pd.testing.assert_frame_equal(d.df, d._get_df())
pd.testing.assert_frame_equal(d.df[0:3], d._get_df(group='group1'))
def test_get_translated_observations(self, df):
d = Data(
df=df,
col_t='time',
col_obs='ln_death_rate',
col_covs=['covariate_1'],
col_group='group',
obs_space=ln_gaussian_pdf,
)
np.testing.assert_equal(
np.exp(d.df[d.col_obs])[0:3],
d._get_translated_observations(group='group1', space=gaussian_pdf))
def test_data(self, df):
d = Data(df=df,
col_t='time',
col_obs='ln_death_rate',
col_covs=['covariate_1'],
col_group='group',
obs_space=ln_gaussian_pdf,
obs_se_func=lambda x: x**(-2.))
assert len(d.df == 6)
np.testing.assert_equal(d.df.time, np.array([1, 4, 5, 1, 6, 7]))
np.testing.assert_equal(d.df.group,
np.repeat(['group1', 'group2'], repeats=3))
np.testing.assert_equal(d.df.obs_se,
np.array([1, 4, 5, 1, 6, 7])**(-2.))
'independent_var': independent_var,
'measurement_value': measurement_value,
'measurement_std': measurement_std,
'cov_one': cov_one,
'social_distance': social_distance,
'data_group': data_group,
}
data_frame = pandas.DataFrame(data_dict)
# ------------------------------------------------------------------------
# curve_model
data = Data(df=data_frame,
col_t='independent_var',
col_obs='measurement_value',
col_covs=['cov_one', 'social_distance'],
col_group='data_group',
obs_space=gaussian_cdf,
col_obs_se='measurement_std')
a_intercept = Variable(covariate='cov_one',
var_link_fun=identity_fun,
fe_init=a_true / 3,
re_init=0.0,
fe_bounds=[-numpy.inf, numpy.inf],
re_bounds=[0.0, 0.0])
b_intercept = Variable(covariate='cov_one',
var_link_fun=identity_fun,
fe_init=b_true / 3,
re_init=0.0,
'group_name': group_names[i],
'obs': group_sim,
'time': time
}))
group_data = pd.concat(groups)
group_data['intercept'] = 1
return group_data
df = simulate_data()
# Use the `Data` object to store information about what the columns represent
data = Data(df=df,
col_t='time',
col_obs='obs',
col_covs=['intercept'],
obs_space=ln_gaussian_pdf,
col_group='group_name',
obs_se_func=lambda x: 1)
""" ```
### Create a Parameter Set
We need to define variables, parameters that use those
variables (in this case there are only intercepts -- and one covariate per parameter -- so the variables
are effectively the same as the parameters), and a parameter set that collects all of that information into one
object.
```python """
alpha_fe = Variable(covariate='intercept',
var_link_fun=lambda x: x,
fe_init=fe_mean[0],
re_init=0.,
data_dict = {
'independent_var': independent_var,
'measurement_value': measurement_value,
'measurement_std': measurement_std,
'constant_one': constant_one,
'data_group': data_group,
}
data_frame = pandas.DataFrame(data_dict)
# ------------------------------------------------------------------------
# curve_model
data = Data(df=data_frame,
col_t='independent_var',
col_obs='measurement_value',
col_covs=num_params * ['constant_one'],
col_group='data_group',
obs_space=expit,
col_obs_se='measurement_std')
a_intercept = Variable(covariate='constant_one',
var_link_fun=lambda x: x,
fe_init=a_true[0] / 3.0,
re_init=0.0,
re_zero_sum_std=abs(a_true[0]) / 100.0,
fe_gprior=[a_true[0], abs(a_true[0]) / 100.0],
fe_bounds=[0.0, numpy.inf],
re_bounds=[-2.0, 2.0])
b_intercept = Variable(covariate='constant_one',
var_link_fun=lambda x: x,