def posterior_predictive(self, target_names=None, data={}):
if self.inference_method is None:
raise RuntimeError(
"posterior_preductive cannot be used before using the fit function."
)
# only non-observed variables can be in target_names
if target_names is None:
target_names = [
name for name in self.vars.keys() if name in self.observed_vars
]
else:
if any(var not in self.observed_vars for var in target_names):
raise ValueError(
"target_names must correspond to observed variables during the inference: \
{}".format(self.observed_vars))
# posterior_predictive uses pmodel variables, but global hidden (parameters) intercepted with qmodel variables.
return Query(
self.inference_method.expanded_variables["p"],
target_names,
data,
enable_interceptor_variables=(
# just interested in intercept the global parameters, not the local hidden
self.inference_method.get_interceptable_condition_variables()
[0],
None))
def posterior_predictive(self, target_names=None, data={}):
# posterior_predictive uses pmodel variables, but global hidden (parameters) intercepted with qmodel variables.
return Query(
self.expanded_variables["p"],
target_names,
data,
# just interested in intercept the global parameters, not the local hidden
enable_interceptor_variables=(self.enable_interceptor_global,
None))
def posterior_predictive(self, target_names=None, data={}):
# posterior_predictive uses pmodel variables, but global hidden (parameters) intercepted with qmodel variables.
expanded_data = {
**data,
**(util.runtime.try_run({
k: v.sample()
for k, v in self.states.items() if k not in data
}))
}
return Query(self.pmodel.vars, target_names, expanded_data)
def prior(self, target_names=None, data={}, size_datamodel=1):
if size_datamodel > 1:
variables, _ = self.expand_model(size_datamodel)
elif size_datamodel == 1:
variables = self.vars
else:
raise ValueError(
"size_datamodel must be greater than 0 but it is {}".format(
size_datamodel))
return Query(variables, target_names, data)
def posterior(self, target_names=None, data={}):
if self.inference_method is None:
raise RuntimeError(
"posterior cannot be used before using the fit function.")
# only non-observed variables can be in target_names
if target_names is None:
target_names = [
name for name in self.vars.keys()
if name not in self.observed_vars
]
else:
if any(var in self.observed_vars for var in target_names):
raise ValueError(
"target_names must correspond to not observed variables during the inference: \
{}".format([
v for v in self.vars.keys()
if v not in self.observed_vars
]))
return Query(self.inference_method.expanded_variables["q"],
target_names, data)
def posterior(self, target_names=None, data={}):
return Query(self.states, target_names, data)
def posterior(self, target_names=None, data={}):
return Query(self.expanded_variables["q"], target_names, data)