# pylint: disable=no-name-in-module
# pylint: disable=too-many-lines
import logging
from copy import deepcopy
import numpy as np
import pandas as pd
import pymc3 as pm
from arviz.plots import plot_posterior
from arviz.data import from_dict
from numpy.linalg import matrix_rank
from formulae import design_matrices
from .backends import PyMC3BackEnd
from .defaults import get_default_prior, get_builtin_family
from .families import Family, _extract_family_prior
from .priors import Prior, PriorScaler, PriorScalerMLE
from .terms import ResponseTerm, Term, GroupSpecificTerm
from .utils import listify, link_match_family
from .version import __version__
_log = logging.getLogger("bambi")
[docs]class Model:
"""Specification of model class.
Parameters
----------
formula : str
A model description written using the formula syntax from the ``formulae`` library.
data : pandas.DataFrame or str
The dataset to use. Either a pandas ``DataFrame``, or the name of the file containing
the data, which will be passed to ``pd.read_csv()``.
family : str or bambi.families.Family
A specification of the model family (analogous to the family object in R). Either
a string, or an instance of class ``bambi.families.Family``. If a string is passed, a
family with the corresponding name must be defined in the defaults loaded at ``Model``
initialization. Valid pre-defined families are ``"bernoulli"``, ``"beta"``,
``"binomial"``, ``"gamma"``, ``"gaussian"``, ``"negativebinomial"``, ``"poisson"``, ``"t"``,
and ``"wald"``. Defaults to ``"gaussian"``.
priors : dict
Optional specification of priors for one or more terms. A dictionary where the keys are
the names of terms in the model, "common" or "group_specific" and the values are
instances of class ``Prior`` when ``automatic_priors`` is ``"default"``, or either
``Prior``, ``int``, ``float``, or ``str`` when ``automatic_priors`` is ``"mle"``.
link : str
The name of the link function to use. Valid names are ``"cloglog"``, ``"identity"``,
``"inverse_squared"``, ``"inverse"``, ``"log"``, ``"logit"``, and ``"probit"``. Not all
the link functions can be used with all the families. See TODO.
categorical : str or list
The names of any variables to treat as categorical. Can be either a single variable
name, or a list of names. If categorical is ``None``, the data type of the columns in
the ``data`` will be used to infer handling. In cases where numeric columns are
to be treated as categoricals (e.g., group specific factors coded as numerical IDs),
explicitly passing variable names via this argument is recommended.
potentials : A list of 2-tuples.
Optional specification of potentials. A potential is an arbitrary expression added to the
likelihood, this is generally useful to add constrains to models, that are difficult to
express otherwise. The first term of a 2-tuple is the name of a variable in the model, the
second a lambda function expressing the desired constraint.
If a constraint involves n variables, you can pass n 2-tuples or pass a tuple which first
element is a n-tuple and second element is a lambda function with n arguments. The number
and order of the lambda function has to match the number and order of the variables names.
dropna : bool
When ``True``, rows with any missing values in either the predictors or outcome are
automatically dropped from the dataset in a listwise manner.
auto_scale : bool
If ``True`` (default), priors are automatically rescaled to the data
(to be weakly informative) any time default priors are used. Note that any priors
explicitly set by the user will always take precedence over default priors.
automatic_priors: str
An optional specification to compute/scale automatic priors. ``"default"`` means to use
a method inspired on the R rstanarm library. ``"mle"`` means to use old default priors in
Bambi that rely on maximum likelihood estimations obtained via the statsmodels library.
noncentered : bool
If ``True`` (default), uses a non-centered parameterization for normal hyperpriors on
grouped parameters. If ``False``, naive (centered) parameterization is used.
priors_cor : dict
An optional value for eta in the LKJ prior for the correlation matrix of group-specific
terms. Keys in the dictionary indicate the groups, and values indicate the value of eta.
This is a very experimental feature. Defaults to ``None``, which means priors for the
group-specific terms are independent.
taylor : int
Order of Taylor expansion to use in approximate variance when constructing the default
priors when ``automatic_priors`` is ``"mle"``. Should be between 1 and 13. Lower values are
less accurate, tending to undershoot the correct prior width, but are faster to compute and
more stable. Odd-numbered values tend to work better. Defaults to 5 for Normal models and
1 for non-Normal models. Values higher than the defaults are generally not recommended as
they can be unstable.
"""
# pylint: disable=too-many-instance-attributes
def __init__(
self,
formula=None,
data=None,
family="gaussian",
priors=None,
link=None,
categorical=None,
potentials=None,
dropna=False,
auto_scale=True,
automatic_priors="default",
noncentered=True,
priors_cor=None,
taylor=None,
):
# attributes that are set later
self.terms = {}
self.built = False # build()
self._backend_name = None
# build() will loop over this, calling _set_priors()
self._added_priors = {}
self._design = None
self.formula = None
self.response = None # _add_response()
self.family = None # _add_response()
self.backend = None # _set_backend()
self.priors_cor = {} # _add_priors_cor()
self.auto_scale = auto_scale
self.dropna = dropna
self.taylor = taylor
self.noncentered = noncentered
self.potentials = potentials
# Read and clean data
if isinstance(data, str):
data = pd.read_csv(data, sep=None, engine="python")
elif not isinstance(data, pd.DataFrame):
raise ValueError("'data' must be a string with a path to a .csv or a pandas DataFrame.")
# To avoid SettingWithCopyWarning when converting object columns to category
data._is_copy = False
# Object columns converted to category by default.
obj_cols = data.select_dtypes(["object"]).columns
data[obj_cols] = data[obj_cols].apply(lambda x: x.astype("category"))
# Explicitly convert columns to category if desired--though this
# can also be done within the formula using C().
if categorical is not None:
data = data.copy()
cats = listify(categorical)
data[cats] = data[cats].apply(lambda x: x.astype("category"))
self.data = data
# Handle priors
if priors is None:
priors = {}
else:
priors = deepcopy(priors)
self.automatic_priors = automatic_priors
# Obtain design matrices and related objects.
if formula is None:
raise ValueError("Can't instantiate a model without a model formula.")
na_action = "drop" if dropna else "error"
self.formula = formula
self._design = design_matrices(formula, data, na_action, eval_env=1)
if self._design.response is None:
raise ValueError(
"No outcome variable is set! "
"Please specify an outcome variable using the formula interface."
)
family_prior = _extract_family_prior(family, priors)
if family_prior and self._design.common:
conflicts = [name for name in family_prior if name in self._design.common.terms_info]
if conflicts:
raise ValueError(
f"The prior name for {', '.join(conflicts)} conflicts with the name of a "
"parameter in the response distribution.\n"
"Please rename the term(s) to prevent an unexpected behaviour."
)
self._add_response(self._design.response, family, link, family_prior)
if self._design.common:
self._add_common(self._design.common, priors)
if self._design.group:
self._add_group_specific(self._design.group, priors)
if priors_cor:
self._add_priors_cor(priors_cor)
# Build priors
self._build_priors()
[docs] def fit(
self,
draws=1000,
tune=1000,
discard_tuned_samples=True,
omit_offsets=True,
method="mcmc",
init="auto",
n_init=50000,
chains=None,
cores=None,
random_seed=None,
**kwargs,
):
"""Fit the model using PyMC3.
Parameters
----------
draws: int
The number of samples to draw from the posterior distribution. Defaults to 1000.
tune : int
Number of iterations to tune. Defaults to 1000. Samplers adjust the step sizes,
scalings or similar during tuning. These tuning samples are be drawn in addition to the
number specified in the ``draws`` argument, and will be discarded unless
``discard_tuned_samples`` is set to ``False``.
discard_tuned_samples : bool
Whether to discard posterior samples of the tune interval. Defaults to ``True``.
omit_offsets: bool
Omits offset terms in the ``InferenceData`` object when the model includes
group specific effects. Defaults to ``True``.
method: str
The method to use for fitting the model. By default, ``"mcmc"``. This automatically
assigns a MCMC method best suited for each kind of variables, like NUTS for continuous
variables and Metropolis for non-binary discrete ones. Alternatively, ``"advi"``, in
which case the model will be fitted using automatic differentiation variational
inference as implemented in PyMC3.
Finally, ``"laplace"``, in which case a Laplace approximation is used and is not
recommended other than for pedagogical use.
init: str
Initialization method. Defaults to ``"auto"``. The available methods are:
* auto: Use ``"jitter+adapt_diag"`` and if this method fails it uses ``"adapt_diag"``.
* adapt_diag: Start with a identity mass matrix and then adapt a diagonal based on the
variance of the tuning samples. All chains use the test value (usually the prior mean)
as starting point.
* jitter+adapt_diag: Same as ``"adapt_diag"``, but use test value plus a uniform jitter
in [-1, 1] as starting point in each chain.
* advi+adapt_diag: Run ADVI and then adapt the resulting diagonal mass matrix based on
the sample variance of the tuning samples.
* advi+adapt_diag_grad: Run ADVI and then adapt the resulting diagonal mass matrix based
on the variance of the gradients during tuning. This is **experimental** and might be
removed in a future release.
* advi: Run ADVI to estimate posterior mean and diagonal mass matrix.
* advi_map: Initialize ADVI with MAP and use MAP as starting point.
* map: Use the MAP as starting point. This is strongly discouraged.
* adapt_full: Adapt a dense mass matrix using the sample covariances. All chains use the
test value (usually the prior mean) as starting point.
* jitter+adapt_full: Same as ``"adapt_full"``, but use test value plus a uniform jitter
in [-1, 1] as starting point in each chain.
n_init: int
Number of initialization iterations. Only works for ``"advi"`` init methods.
chains: int
The number of chains to sample. Running independent chains is important for some
convergence statistics and can also reveal multiple modes in the posterior. If ``None``,
then set to either ``cores`` or 2, whichever is larger.
cores : int
The number of chains to run in parallel. If ``None``, it is equal to the number of CPUs
in the system unless there are more than 4 CPUs, in which case it is set to 4.
random_seed : int or list of ints
A list is accepted if cores is greater than one.
**kwargs:
For other kwargs see the documentation for ``pymc3.sample()``.
Returns
-------
An ArviZ ``InferenceData`` instance.
"""
if not self.built:
self.build()
# Tell user which event is being modeled
if self.family.name == "bernoulli":
_log.info(
"Modeling the probability that %s==%s",
self.response.name,
str(self.response.success_event),
)
return self.backend.run(
draws=draws,
tune=tune,
discard_tuned_samples=discard_tuned_samples,
omit_offsets=omit_offsets,
method=method,
init=init,
n_init=n_init,
chains=chains,
cores=cores,
random_seed=random_seed,
**kwargs,
)
[docs] def build(self):
"""Set up the model for sampling/fitting.
Performs any steps that require access to all model terms (e.g., scaling priors
on each term), then calls the backend's ``build()`` method.
"""
self.backend = PyMC3BackEnd()
self.backend.build(self)
self.built = True
[docs] def set_priors(self, priors=None, common=None, group_specific=None):
"""Set priors for one or more existing terms.
Parameters
----------
priors : dict
Dictionary of priors to update. Keys are names of terms to update; values are the new
priors (either a ``Prior`` instance, or an int or float that scales the default priors).
Note that a tuple can be passed as the key, in which case the same prior will be applied
to all terms named in the tuple.
common : Prior, int, float or str
A prior specification to apply to all common terms included in the model.
group_specific : Prior, int, float or str
A prior specification to apply to all group specific terms included in the model.
"""
# save arguments to pass to _set_priors() at build time
kwargs = dict(zip(["priors", "common", "group_specific"], [priors, common, group_specific]))
self._added_priors.update(kwargs)
# After updating, we need to rebuild priors.
# There is redundancy here, so there's place for performance improvements.
self._build_priors()
self.built = False
def _build_priors(self):
"""Carry out all operations related to the construction and/or scaling of priors."""
# Set custom priors that have been passed via `Model.set_priors()`
self._set_priors(**self._added_priors)
# Prepare all priors
for term in self.terms.values():
if term.group_specific:
type_ = "group_specific"
elif term.type == "intercept":
type_ = "intercept"
else:
type_ = "common"
term.prior = self._prepare_prior(term.prior, type_)
# Scale priors if there is at least one term in the model and auto_scale is True
if self.terms and self.auto_scale:
method = self.automatic_priors
if method == "default":
scaler = PriorScaler(self)
elif method == "mle":
if self.taylor is not None:
taylor = self.taylor
else:
taylor = 5 if self.family.name == "gaussian" else 1
scaler = PriorScalerMLE(self, taylor=taylor)
else:
raise ValueError(
f"{method} is not a valid method for default priors." "Use 'default' or 'mle'."
)
self.scaler = scaler
self.scaler.scale()
def _set_priors(self, priors=None, common=None, group_specific=None):
"""Internal version of ``set_priors()``, with same arguments.
Runs during ``Model._build_priors()``.
"""
# First, it constructs a `targets` dict where it store key-value (name-prior) pairs that
# are going to be updated. Finally, the update is done in the last for loop in this method.
targets = {}
if common is not None:
targets.update({name: common for name in self.common_terms.keys()})
if group_specific is not None:
targets.update({name: group_specific for name in self.group_specific_terms.keys()})
if priors is not None:
# Prepare priors for response auxiliary parameters
family_prior = _extract_family_prior(self.family, priors)
if family_prior:
for prior in family_prior.values():
prior.auto_scale = False
self.family.likelihood.priors.update(family_prior)
# Prepare priors for explanatory terms.
for names, prior in priors.items():
# In case we have tuple-keys, we loop throuh each of them.
for name in listify(names):
if name not in list(self.terms.keys()):
raise ValueError(f"No terms in model match {name}.")
targets[name] = prior
# Set priors for explanatory terms.
for name, prior in targets.items():
self.terms[name].prior = prior
def _prepare_prior(self, prior, type_):
"""Helper function to correctly set default priors, auto scaling, etc.
Parameters
----------
prior : Prior, float, or None.
type_ : string
Accepted values are: ``"intercept"``, ``"common"``, or ``"group_specific"``.
"""
if prior is None and not self.auto_scale:
prior = get_default_prior(type_ + "_flat")
if isinstance(prior, Prior):
prior.auto_scale = False
else:
_scale = prior
prior = get_default_prior(type_)
prior.scale = _scale
return prior
def _add_response(self, response, family="gaussian", link=None, priors=None):
"""Add a response (or outcome/dependent) variable to the model.
Parameters
----------
response : formulae.ResponseVector
An instance of ``formulae.ResponseVector`` as returned by
``formulae.design_matrices()``.
family : str or bambi.families.Family
A specification of the model family (analogous to the family object in R). Either a
string, or an instance of class ``families.Family``. If a string is passed, a family
with the corresponding name must be defined in the defaults loaded at model
initialization. Valid pre-defined families are ``"bernoulli"``, ``"beta"``,
``"binomial"``, ``"gamma"``, ``"gaussian"``, ``"negativebinomial"``, ``"poisson"``,
``"t"``, and ``"wald"``. Defaults to ``"gaussian"``.
link : str
The name of the link function to use. Valid names are ``"cloglog"``, ``"identity"``,
``"inverse_squared"``, ``"inverse"``, ``"log"``, ``"logit"``, and ``"probit"``. Not all
the link functions can be used with all the families. See TODO.
priors : dict
Optional dictionary with specification of priors for the parameters in the family of
the response. Keys are names of other parameters than the mean in the family
(i.e. they cannot be equal to ``family.parent``) and values can be an instance of class
``Prior``, a numeric value, or a string describing the width. In the numeric case,
the distribution specified in the defaults will be used, and the passed value will be
used to scale the appropriate variance parameter. Strings, which are only available when
``automatic_priors`` is ``"mle"``and can be one of ``"wide"``, ``"narrow"``,
``"medium"``, or ``"superwide"``), predefined values will be used.
"""
if isinstance(family, str):
family = get_builtin_family(family)
elif not isinstance(family, Family):
raise ValueError("family must be a string or a Family object.")
# Override family's link if another is explicitly passed
if link is not None:
if link_match_family(link, family.name):
family._set_link(link) # pylint: disable=protected-access
else:
raise ValueError(f"Link '{link}'' cannot be used with family '{family.name}'")
# Update auxiliary parameters
if priors:
for prior in priors.values():
if isinstance(prior, Prior):
prior.auto_scale = False
family.likelihood.priors.update(priors)
if response.success is not None and family.name != "bernoulli":
raise ValueError("Index notation for response is only available for 'bernoulli' family")
self.family = family
self.response = ResponseTerm(response, family)
self.built = False
def _add_common(self, common, priors):
"""Add common (a.k.a. fixed) terms to the model.
Parameters
----------
common : formulae.CommonEffectsMatrix
Representation of the design matrix for the common effects of a model. It contains all
the necessary information to build the ``Term`` objects associated with each common
term in the model.
priors : dict
Optional specification of priors for one or more terms. A dictionary where the keys are
any of the names of the common terms in the model or ``"common"`` and the values are
either instances of class ``Prior`` or ``int``, ``float``, or ``str`` that specify the
width of the priors on a standardized scale.
"""
if matrix_rank(common.design_matrix) < common.design_matrix.shape[1]:
raise ValueError(
"Design matrix for common effects is not full-rank. "
"Bambi does not support sparse settings yet."
)
for name, term in common.terms_info.items():
data = common[name]
prior = priors.pop(name, priors.get("common", None))
if isinstance(prior, Prior):
any_hyperprior = any(isinstance(x, Prior) for x in prior.args.values())
if any_hyperprior:
raise ValueError(
f"Trying to set hyperprior on '{name}'. "
"Can't set a hyperprior on common effects."
)
self.terms[name] = Term(name, term, data, prior)
def _add_group_specific(self, group, priors):
"""Add group-specific (or random) terms to the model.
Parameters
----------
group : formulae.GroupEffectsMatrix
Representation of the design matrix for the group specific effects of a model. It
contains all the necessary information to build the ``GroupSpecificTerm`` objects
associated with each group-specific term in the model.
priors : dict
Optional specification of priors for one or more terms. A dictionary where the keys are
any of the names of the group-specific terms in the model or ``"group_specific"`` and
the values are either instances of class ``Prior`` or ``int``, ``float``, or ``str``
that specify the width of the priors on a standardized scale.
"""
for name, term in group.terms_info.items():
data = group[name]
prior = priors.pop(name, priors.get("group_specific", None))
self.terms[name] = GroupSpecificTerm(name, term, data, prior)
def _add_priors_cor(self, priors):
# priors: dictionary. names are groups, values are the "eta" in the lkj prior
groups = self._get_group_specific_groups()
for group in groups:
if group in priors:
self.priors_cor[group] = priors[group]
else:
raise KeyError(f"The name {group} is not a group in any group-specific term.")
[docs] def plot_priors(
self,
draws=5000,
var_names=None,
random_seed=None,
figsize=None,
textsize=None,
hdi_prob=None,
round_to=2,
point_estimate="mean",
kind="kde",
bins=None,
omit_offsets=True,
omit_group_specific=True,
ax=None,
):
"""
Samples from the prior distribution and plots its marginals.
Parameters
----------
draws : int
Number of draws to sample from the prior predictive distribution. Defaults to 5000.
var_names : str or list
A list of names of variables for which to compute the posterior predictive
distribution. Defaults to ``None`` which means to include both observed and
unobserved RVs.
random_seed : int
Seed for the random number generator.
figsize: tuple
Figure size. If ``None`` it will be defined automatically.
textsize: float
Text size scaling factor for labels, titles and lines. If ``None`` it will be
autoscaled based on ``figsize``.
hdi_prob: float or str
Plots highest density interval for chosen percentage of density.
Use ``"hide"`` to hide the highest density interval. Defaults to 0.94.
round_to: int
Controls formatting of floats. Defaults to 2 or the integer part, whichever is bigger.
point_estimate: str
Plot point estimate per variable. Values should be ``"mean"``, ``"median"``, ``"mode"``
or ``None``. Defaults to ``"auto"`` i.e. it falls back to default set in
ArviZ's rcParams.
kind: str
Type of plot to display (``"kde"`` or ``"hist"``) For discrete variables this argument
is ignored and a histogram is always used.
bins: integer or sequence or "auto"
Controls the number of bins, accepts the same keywords ``matplotlib.pyplot.hist()``
does. Only works if ``kind == "hist"``. If ``None`` (default) it will use ``"auto"``
for continuous variables and ``range(xmin, xmax + 1)`` for discrete variables.
omit_offsets: bool
Whether to omit offset terms in the plot. Defaults to ``True``.
omit_group_specific: bool
Whether to omit group specific effects in the plot. Defaults to ``True``.
ax: numpy array-like of matplotlib axes or bokeh figures
A 2D array of locations into which to plot the densities. If not supplied, ArviZ will
create its own array of plot areas (and return it).
**kwargs
Passed as-is to ``matplotlib.pyplot.hist()`` or ``matplotlib.pyplot.plot()`` function
depending on the value of ``kind``.
Returns
-------
axes: matplotlib axes
"""
if not self.built:
raise ValueError(
"Cannot plot priors until model is built!! "
"Call .build() to build the model or .fit() to build and sample from the posterior."
)
unobserved_rvs_names = []
flat_rvs = []
for unobserved in self.backend.model.unobserved_RVs:
if "Flat" in unobserved.__str__():
flat_rvs.append(unobserved.name)
else:
unobserved_rvs_names.append(unobserved.name)
if var_names is None:
var_names = pm.util.get_default_varnames(
unobserved_rvs_names, include_transformed=False
)
else:
flat_rvs = [fv for fv in flat_rvs if fv in var_names]
var_names = [vn for vn in var_names if vn not in flat_rvs]
if flat_rvs:
_log.info(
"Variables %s have flat priors, and hence they are not plotted", ", ".join(flat_rvs)
)
if omit_offsets:
var_names = [name for name in var_names if not name.endswith("_offset")]
if omit_group_specific:
omitted = list(self.group_specific_terms)
var_names = [vn for vn in var_names if vn not in omitted]
axes = None
if var_names:
# Sort variable names so Intercept is in the beginning
if "Intercept" in var_names:
var_names.insert(0, var_names.pop(var_names.index("Intercept")))
pps = self.prior_predictive(draws=draws, var_names=var_names, random_seed=random_seed)
axes = plot_posterior(
pps,
group="prior",
figsize=figsize,
textsize=textsize,
hdi_prob=hdi_prob,
round_to=round_to,
point_estimate=point_estimate,
kind=kind,
bins=bins,
ax=ax,
)
return axes
[docs] def prior_predictive(self, draws=500, var_names=None, omit_offsets=True, random_seed=None):
"""
Generate samples from the prior predictive distribution.
Parameters
----------
draws : int
Number of draws to sample from the prior predictive distribution. Defaults to 500.
var_names : str or list
A list of names of variables for which to compute the prior predictive distribution.
Defaults to ``None`` which means both observed and unobserved RVs.
random_seed : int
Seed for the random number generator.
Returns
-------
InferenceData
``InferenceData`` object with the groups ``prior``, ``prior_predictive`` and
``observed_data``.
"""
if var_names is None:
variables = self.backend.model.unobserved_RVs + self.backend.model.observed_RVs
variables_names = [v.name for v in variables]
var_names = pm.util.get_default_varnames(variables_names, include_transformed=False)
if omit_offsets:
var_names = [name for name in var_names if not name.endswith("_offset")]
pps_ = pm.sample_prior_predictive(
samples=draws, var_names=var_names, model=self.backend.model, random_seed=random_seed
)
# pps_ keys are not in the same order as `var_names` because `var_names` is converted
# to set within pm.sample_prior_predictive()
pps = {}
for name in var_names:
if name in self.terms and self.terms[name].categorical:
pps[name] = pps_[name]
else:
pps[name] = pps_[name].squeeze()
response_name = self.response.name
if response_name in pps:
prior_predictive = {response_name: pps.pop(response_name)[np.newaxis]}
observed_data = {response_name: self.response.data.squeeze()}
else:
prior_predictive = {}
observed_data = {}
prior = {k: v[np.newaxis] for k, v in pps.items()}
coords = {}
dims = {}
for name in var_names:
if name in self.terms:
coords.update(**self.terms[name].pymc_coords)
dims[name] = list(self.terms[name].pymc_coords.keys())
idata = from_dict(
prior_predictive=prior_predictive,
prior=prior,
observed_data=observed_data,
coords=coords,
dims=dims,
attrs={
"inference_library": self.backend.name,
"inference_library_version": self.backend.name,
"modeling_interface": "bambi",
"modeling_interface_version": __version__,
},
)
return idata
# pylint: disable=protected-access
[docs] def predict(self, idata, kind="mean", data=None, draws=None, inplace=True):
"""Predict method for Bambi models
Obtains in-sample and out-of-sample predictions from a fitted Bambi model.
Parameters
----------
idata : InferenceData
The ``InferenceData`` instance returned by ``.fit()``.
kind: str
Indicates the type of prediction required. Can be ``"mean"`` or ``"pps"``. The
first returns draws from the posterior distribution of the mean, while the latter
returns the draws from the posterior predictive distribution
(i.e. the posterior probability distribution for a new observation).
Defaults to ``"mean"``.
data: pandas.DataFrame or None
An optional data frame with values for the predictors that are used to obtain
out-of-sample predictions. If omitted, the original dataset is used.
draws: None
The number of random draws per chain. Only used if ``kind="pps"``. Not recommended
unless more than ndraws times nchains posterior predictive samples are needed.
Defaults to ``None`` which means ndraws times nchains draws are obtained.
inplace: bool
If ``True`` it will modify ``idata`` in-place. Otherwise, it will return a copy of
``idata`` with the predictions added. If ``kind="mean"``, a new variable ending in
``"_mean"`` is added to the ``posterior`` group. If ``kind="pps"``, it appends a
``posterior_predictive`` group to ``idata``. If any of these already exist, it will be
overwritten.
Returns
-------
InferenceData or None
"""
if kind not in ["mean", "pps"]:
raise ValueError("'kind' must be one of 'mean' or 'pps'")
linear_predictor = 0
X = None
Z = None
chain_n = len(idata.posterior["chain"])
draw_n = len(idata.posterior["draw"])
posterior = idata.posterior.stack(sample=["chain", "draw"])
if draws is None:
draws = draw_n
if not inplace:
idata = deepcopy(idata)
in_sample = data is None
# Create design matrices
if self._design.common:
if in_sample:
X = self._design.common.design_matrix
else:
X = self._design.common._evaluate_new_data(data).design_matrix
if self._design.group:
if in_sample:
Z = self._design.group.design_matrix
else:
Z = self._design.group._evaluate_new_data(data).design_matrix
# Obtain posterior and compute linear predictor
if X is not None:
beta_x_list = [np.atleast_2d(posterior[name]) for name in self.common_terms]
if self.intercept_term:
beta_x_list = [np.atleast_2d(posterior["Intercept"])] + beta_x_list
beta_x = np.vstack(beta_x_list)
linear_predictor += np.dot(X, beta_x)
if Z is not None:
beta_z = np.vstack(
[np.atleast_2d(posterior[name]) for name in self.group_specific_terms]
)
linear_predictor += np.dot(Z, beta_z)
# Compute mean prediction
# Transposed so it is (chain, draws)?
mu = self.family.link.linkinv(linear_predictor).T
# Reshape mu
obs_n = mu.size // (chain_n * draw_n)
mu = mu.reshape((chain_n, draw_n, obs_n))
# Predictions for the mean
if kind == "mean":
name = self.response.name + "_mean"
coord_name = name + "_dim_0"
# Drop var/dim if already present
if name in idata.posterior.data_vars:
idata.posterior = idata.posterior.drop_vars(name).drop_dims(coord_name)
idata.posterior[name] = (("chain", "draw", coord_name), mu)
idata.posterior = idata.posterior.assign_coords({coord_name: list(range(obs_n))})
# Compute posterior predictive distribution
else:
# Sample mu values and auxiliary params
if not in_sample and self.family.name == "binomial":
n = self._design.response._evaluate_new_data(data)
pps = self.family.likelihood.pps(self, idata.posterior, mu, draws, draw_n, trials=n)
else:
pps = self.family.likelihood.pps(self, idata.posterior, mu, draws, draw_n)
if "posterior_predictive" in idata:
del idata.posterior_predictive
idata.add_groups({"posterior_predictive": {self.response.name: pps}})
getattr(idata, "posterior_predictive").attrs["modeling_interface"] = "bambi"
getattr(idata, "posterior_predictive").attrs["modeling_interface_version"] = __version__
if inplace:
return None
else:
return idata
[docs] def graph(self, formatting="plain", name=None, figsize=None, dpi=300, fmt="png"):
"""
Produce a graphviz Digraph from a built Bambi model.
Requires graphviz, which may be installed most easily with
``conda install -c conda-forge python-graphviz``
Alternatively, you may install the ``graphviz`` binaries yourself, and then
``pip install graphviz`` to get the python bindings.
See http://graphviz.readthedocs.io/en/stable/manual.html for more information.
Parameters
----------
formatting : str
One of ``"plain"`` or ``"plain_with_params"``. Defaults to ``"plain"``.
name : str
Name of the figure to save. Defaults to ``None``, no figure is saved.
figsize : tuple
Maximum width and height of figure in inches. Defaults to ``None``, the figure size is
set automatically. If defined and the drawing is larger than the given size, the drawing
is uniformly scaled down so that it fits within the given size. Only works if ``name``
is not ``None``.
dpi : int
Point per inch of the figure to save.
Defaults to 300. Only works if ``name`` is not ``None``.
fmt : str
Format of the figure to save.
Defaults to ``"png"``. Only works if ``name`` is not ``None``.
Example
--------
>>> model = Model("y ~ x + (1|z)")
>>> model.build()
>>> model.graph()
>>> model = Model("y ~ x + (1|z)")
>>> model.fit()
>>> model.graph()
"""
if self.backend is None:
raise ValueError(
"The model is empty. "
"Are you forgetting to first call .build() or .fit() on the Bambi model?"
)
graphviz = pm.model_to_graphviz(model=self.backend.model, formatting=formatting)
width, height = (None, None) if figsize is None else figsize
if name is not None:
graphviz_ = graphviz.copy()
graphviz_.graph_attr.update(size=f"{width},{height}!")
graphviz_.graph_attr.update(dpi=str(dpi))
graphviz_.render(filename=name, format=fmt, cleanup=True)
return graphviz
def _get_pymc_coords(self):
coords = {}
for term in self.terms.values():
coords.update(**term.pymc_coords)
return coords
def _get_group_specific_groups(self):
groups = {}
for term_name in self.group_specific_terms:
factor = term_name.split("|")[1]
if factor not in groups:
groups[factor] = [term_name]
else:
groups[factor].append(term_name)
return groups
def __str__(self):
priors = ""
priors_common = [f" {t.name} ~ {t.prior}" for t in self.common_terms.values()]
priors_group = [f" {t.name} ~ {t.prior}" for t in self.group_specific_terms.values()]
# Prior for the correlation matrix in group-specific terms
priors_cor = [f" {k} ~ LKJCorr({v})" for k, v in self.priors_cor.items()]
# Priors for auxiliary parameters, e.g., standard deviation in normal linear model
priors_aux = [f" {k} ~ {v}" for k, v in self.family.likelihood.priors.items()]
if self.intercept_term:
t = self.intercept_term
priors_common = [f" {t.name} ~ {t.prior}"] + priors_common
if priors_common:
priors += "\n".join([" Common-level effects", *priors_common]) + "\n\n"
if priors_group:
priors += "\n".join([" Group-level effects", *priors_group]) + "\n\n"
if priors_cor:
priors += "\n".join([" Group-level correlation", *priors_cor]) + "\n\n"
if priors_aux:
priors += "\n".join([" Auxiliary parameters", *priors_aux]) + "\n\n"
str_list = [
f"Formula: {self.formula}",
f"Family name: {self.family.name.capitalize()}",
f"Link: {self.family.link.name}",
f"Observations: {self.response.data.shape[0]}",
"Priors:",
priors,
]
if self.backend and self.backend.fit:
extra_foot = "------\n"
extra_foot += "* To see a plot of the priors call the .plot_priors() method.\n"
extra_foot += "* To see a summary or plot of the posterior pass the object returned"
extra_foot += " by .fit() to az.summary() or az.plot_trace()\n"
str_list += [extra_foot]
return "\n".join(str_list)
def __repr__(self):
return self.__str__()
@property
def term_names(self):
"""Return names of all terms in order of addition to model."""
return list(self.terms.keys())
@property
def common_terms(self):
"""Return dict of all and only common effects in model."""
return {
k: v for (k, v) in self.terms.items() if not v.group_specific and v.type != "intercept"
}
@property
def group_specific_terms(self):
"""Return dict of all and only group specific effects in model."""
return {k: v for (k, v) in self.terms.items() if v.group_specific}
@property
def intercept_term(self):
"""Return the intercept term"""
term = [v for v in self.terms.values() if not v.group_specific and v.type == "intercept"]
if term:
return term[0]
else:
return None