def test_drop_nan(self):
nan = float('nan') # NaN
array_with_nan_1d = np.array([nan, nan, 1, 2, 3])
returned_array_1d = util.drop_nan(array_with_nan_1d)
self.assertTrue(all(returned_array_1d == np.array([1, 2, 3])))
array_with_nan_2d = np.array([[nan], [nan], [1], [2], [3]])
returned_array_2d = util.drop_nan(array_with_nan_2d)
print(returned_array_2d)
self.assertTrue(all(returned_array_2d == np.array([[1], [2], [3]])))
array_without_nan_1d = np.array([1, 2])
returned_array = util.drop_nan(array_without_nan_1d)
self.assertTrue(all(returned_array == np.array([1, 2])))
def test_drop_nan(self):
nan = float('nan') # NaN
array_with_nan_1d = np.array([nan, nan, 1, 2, 3])
returned_array_1d = util.drop_nan(array_with_nan_1d)
self.assertTrue(all(returned_array_1d == np.array([1, 2, 3])))
array_with_nan_2d = np.array([[nan], [nan], [1], [2], [3]])
returned_array_2d = util.drop_nan(array_with_nan_2d)
print(returned_array_2d)
self.assertTrue(all(returned_array_2d == np.array([[1], [2], [3]])))
array_without_nan_1d = np.array([1, 2])
returned_array = util.drop_nan(array_without_nan_1d)
self.assertTrue(all(returned_array == np.array([1, 2])))
def _bayes_sampling(x,
y,
distribution='normal',
num_iters=25000,
inference="sampling"):
""" Helper function for bayesian sampling.
:param x: sample of a treatment group
:type x: pd.Series or list (array-like)
:param y: sample of a control group
:type y: pd.Series or list (array-like)
:param distribution: name of the KPI distribution model, which assumes a Stan model file with the same name exists
:type distribution: str
:param num_iters: number of iterations of sampling
:type num_iters: int
:param inference: 'sampling' for MCMC sampling method or 'variational' for variational inference
:type inference: str
:return: the posterior samples, sample size of x, sample size of y, absolute mean of x, absolute mean of y
:rtype: tuple[array-like, array-like, array-like, float, float]
"""
# Checking if data was provided and it has correct format
if x is None or y is None:
raise ValueError('Please provide two non-empty samples.')
if not isinstance(x, pd.Series) and not isinstance(
x, np.ndarray) and not isinstance(x, list):
raise TypeError('Please provide samples of type Series or list.')
if type(x) != type(y):
raise TypeError('Please provide samples of the same type.')
logger.info(
"Started running bayesian inference with {} procedure, treatment group of size {}, "
"control group of size {}, {} distribution.".format(
inference, len(x), len(y), distribution, inference))
# Coercing missing values to right format
_x = np.array(x, dtype=float)
_y = np.array(y, dtype=float)
_x = drop_nan(_x)
_y = drop_nan(_y)
key = (str(_x), str(_y), num_iters, inference)
if cache_sampling_results and key in sampling_results:
return sampling_results[key]
mu_x = np.nanmean(_x)
mu_y = np.nanmean(_y)
n_x = statx.sample_size(_x)
n_y = statx.sample_size(_y)
if distribution == 'normal':
fit_data = {'Nc': n_y, 'Nt': n_x, 'x': _x, 'y': _y}
elif distribution == 'poisson':
fit_data = {
'Nc': n_y,
'Nt': n_x,
'x': _x.astype(int),
'y': _y.astype(int)
}
else:
raise NotImplementedError
model_file = __location__ + '/../models/' + distribution + '_kpi.stan'
sm = get_or_compile_stan_model(model_file, distribution)
if inference == "sampling":
fit = sm.sampling(data=fit_data,
iter=num_iters,
chains=4,
n_jobs=1,
seed=1,
control={
'stepsize': 0.01,
'adapt_delta': 0.99
})
traces = fit.extract()
elif inference == "variational":
results_dict = sm.vb(data=fit_data, iter=10000)
traces = {}
for i in range(len(results_dict['sampler_param_names'])):
para_name = results_dict['sampler_param_names'][i]
para_values = np.array(results_dict['sampler_params'][i])
traces[para_name] = para_values
if cache_sampling_results:
sampling_results[key] = (traces, n_x, n_y, mu_x, mu_y)
logger.info(
"Finished running bayesian inference with {} procedure, treatment group of size {}, "
"control group of size {}, {} distribution.".format(
inference, len(x), len(y), distribution))
return traces, n_x, n_y, mu_x, mu_y
def _bayes_sampling(x, y, distribution='normal', num_iters=25000, inference="sampling"):
""" Helper function for bayesian sampling.
:param x: sample of a treatment group
:type x: pd.Series or list (array-like)
:param y: sample of a control group
:type y: pd.Series or list (array-like)
:param distribution: name of the KPI distribution model, which assumes a Stan model file with the same name exists
:type distribution: str
:param num_iters: number of iterations of sampling
:type num_iters: int
:param inference: 'sampling' for MCMC sampling method or 'variational' for variational inference
:type inference: str
:return: the posterior samples, sample size of x, sample size of y, absolute mean of x, absolute mean of y
:rtype: tuple[array-like, array-like, array-like, float, float]
"""
# Checking if data was provided and it has correct format
if x is None or y is None:
raise ValueError('Please provide two non-empty samples.')
if not isinstance(x, pd.Series) and not isinstance(x, np.ndarray) and not isinstance(x, list):
raise TypeError('Please provide samples of type Series or list.')
if type(x) != type(y):
raise TypeError('Please provide samples of the same type.')
logger.info("Started running bayesian inference with {} procedure, treatment group of size {}, "
"control group of size {}, {} distribution.".format(inference, len(x), len(y), distribution, inference))
# Coercing missing values to right format
_x = np.array(x, dtype=float)
_y = np.array(y, dtype=float)
_x = drop_nan(_x)
_y = drop_nan(_y)
key = (str(_x), str(_y), num_iters, inference)
if cache_sampling_results and key in sampling_results:
return sampling_results[key]
mu_x = np.nanmean(_x)
mu_y = np.nanmean(_y)
n_x = statx.sample_size(_x)
n_y = statx.sample_size(_y)
if distribution == 'normal':
fit_data = {'Nc': n_y,
'Nt': n_x,
'x': _x,
'y': _y}
elif distribution == 'poisson':
fit_data = {'Nc': n_y,
'Nt': n_x,
'x': _x.astype(int),
'y': _y.astype(int)}
else:
raise NotImplementedError
model_file = __location__ + '/../models/' + distribution + '_kpi.stan'
sm = get_or_compile_stan_model(model_file, distribution)
if inference == "sampling":
fit = sm.sampling(data=fit_data, iter=num_iters, chains=4, n_jobs=1, seed=1,
control={'stepsize': 0.01, 'adapt_delta': 0.99})
traces = fit.extract()
elif inference == "variational":
results_dict = sm.vb(data=fit_data, iter=10000)
traces = {}
for i in range(len(results_dict['sampler_param_names'])):
para_name = results_dict['sampler_param_names'][i]
para_values = np.array(results_dict['sampler_params'][i])
traces[para_name] = para_values
if cache_sampling_results:
sampling_results[key] = (traces, n_x, n_y, mu_x, mu_y)
logger.info("Finished running bayesian inference with {} procedure, treatment group of size {}, "
"control group of size {}, {} distribution.".format(inference, len(x), len(y), distribution))
return traces, n_x, n_y, mu_x, mu_y
def _bayes_sampling(x,
y,
distribution='normal',
num_iters=25000,
inference="sampling"):
"""
Helper function.
Args:
x (array_like): sample of a treatment group
y (array_like): sample of a control group
distribution: name of the KPI distribution model, which assumes a
Stan model file with the same name exists
num_iters: number of iterations of sampling
Returns:
tuple:
- the posterior samples
- sample size of x
- sample size of y
- absolute mean of x
- absolute mean of y
"""
# Checking if data was provided
if x is None or y is None:
raise ValueError('Please provide two non-None samples.')
# Coercing missing values to right format
_x = np.array(x, dtype=float)
_y = np.array(y, dtype=float)
_x = drop_nan(_x)
_y = drop_nan(_y)
key = (str(_x), str(_y), num_iters, inference)
if cache_sampling_results and key in sampling_results:
return sampling_results[key]
mu_x = np.nanmean(_x)
mu_y = np.nanmean(_y)
n_x = statx.sample_size(_x)
n_y = statx.sample_size(_y)
if distribution == 'normal':
fit_data = {'Nc': n_y, 'Nt': n_x, 'x': _x, 'y': _y}
elif distribution == 'poisson':
fit_data = {
'Nc': n_y,
'Nt': n_x,
'x': _x.astype(int),
'y': _y.astype(int)
}
else:
raise NotImplementedError
model_file = __location__ + '/../models/' + distribution + '_kpi.stan'
sm = get_or_compile_stan_model(model_file, distribution)
if inference == "sampling":
fit = sm.sampling(data=fit_data,
iter=num_iters,
chains=4,
n_jobs=1,
seed=1,
control={
'stepsize': 0.01,
'adapt_delta': 0.99
})
traces = fit.extract()
elif inference == "variational":
results_dict = sm.vb(data=fit_data, iter=10000)
traces = {}
for i in range(len(results_dict['sampler_param_names'])):
para_name = results_dict['sampler_param_names'][i]
para_values = np.array(results_dict['sampler_params'][i])
traces[para_name] = para_values
if cache_sampling_results:
sampling_results[key] = (traces, n_x, n_y, mu_x, mu_y)
return traces, n_x, n_y, mu_x, mu_y