multiple_inference.bayes.base#

Base classes for Bayesian analysis.

Classes

BayesBase(mean, cov[, X, endog_names, ...])

Mixin for Bayesian models.

BayesResults(*args[, n_samples])

Results of Bayesian analysis.
class multiple_inference.bayes.base.BayesBase(mean: Sequence[float], cov: ndarray, X: ndarray | None = None, endog_names: str | None = None, exog_names: Sequence[str] | None = None, columns: Sequence[int] | Sequence[str] | Sequence[bool] | None = None, sort: bool = False, random_state: int = 0)[source]#

Mixin for Bayesian models.

Subclasses multiple_inference.base.ModelBase.

get_joint_distribution(columns: Sequence[int] | Sequence[str] | Sequence[bool] | None = None)[source]#

Get the joint posterior distribution.

Parameters:

columns (ColumnsType, optional) – Selected columns. Defaults to None.

Returns:

Joint distribution.

Return type:

rv_like

get_joint_prior(columns: Sequence[int] | Sequence[str] | Sequence[bool] | None = None)[source]#

Get the joint prior distribution.

Parameters:

columns (ColumnsType, optional) – Selected columns. Defaults to None.

Returns:

Joint distribution.

Return type:

rv_like

get_marginal_distribution(column: str | int) → rv_continuous[source]#

Get the marginal posterior distribution of column.

Parameters:

column (ColumnType) – Name or index of the parameter of interest.

Returns:

Posterior distribution.

Return type:

rv_continuous

get_marginal_prior(column: str | int) → rv_continuous[source]#

Get the marginal prior distribution of column.

Parameters:

column (ColumnType) – Name or index of the parameter of interest.

Returns:

Prior distribution

Return type:

rv_continuous

class multiple_inference.bayes.base.BayesResults(*args: Any, n_samples: int = 10000, **kwargs: Any)[source]#

Results of Bayesian analysis.

Inherits from multiple_inference.base.ResultsBase.

Parameters:
distributions#

Marginal posterior distributions.

Type:

List[scipy.stats.norm]

multivariate_distribution#

Joint posterior distribution.

Type:

scipy.stats.multivariate_normal

rank_df#

(n, n) dataframe of probabilities that column i has rank j.

Type:

pd.DataFrame

compute_best_params(n_best_params: int = 1, alpha: float = 0.05, superset: bool = True, criterion: str = 'fwer') → Series[source]#

Compute the set of best (largest) parameters.

Parameters:

  • n_best_params (int, optional) – Number of best parameters. Defaults to 1.

  • alpha (float, optional) – Significance level. If the criterion is “fwer”, alpha is the family-wise error rate. If the criterion is “fdr”, alpha is the false discovery rate. Defaults to 0.05.

  • superset (bool, optional) – Indicates that the returned set should be a superset of the truly best parameters. If False, the returned set should be a subset of the truly best parameters. Defaults to True.

  • criterion (str, optional) – “fwer” to control the family-wise error rate, “fdr” to control the false discovery rate. Defaults to “fwer”.

Raises:

ValueError – criterion must be either “fwer” or “fdr”.

Returns:

Indicates which parameters are in the selected set.

Return type:

pd.Series

expected_wasserstein_distance(mean: Sequence[float] | None = None, cov: ndarray | None = None, **kwargs: Any) → float[source]#

Compute the Wasserstein distance metric.

This method estimates the Wasserstein distance between the observed distribution (a joint normal characterized by mean and cov) and the distribution you would expect to observe according to this model.

Parameters:

  • mean (Numeric1DArray, optional) – (# params,) array of sample conventionally estimated means. If None, use the model’s estimated means. Defaults to None.

  • cov (np.ndarray, optional) – (# params, # params) covaraince matrix for conventionally estimated means. If None, use the model’s estimated covariance matrix. Defaults to None.

  • **kwargs (Any) – Keyword arguments for scipy.stats.wasserstein_distance.

Returns:

Expected Wasserstein distance.

Return type:

float

likelihood(mean: Sequence[float] | None = None, cov: ndarray | None = None) → float[source]#
Parameters:

  • mean (Numeric1DArray, optional) – (# params,) array of sample conventionally estimated means. If None, use the model’s estimated means. Defaults to None.

  • cov (np.ndarray, optional) – (# params, # params) covaraince matrix for conventionally estimated means. If None, use the model’s estimated covariance matrix. Defaults to None.

Returns:

Likelihood.

Return type:

float

line_plot(column: int | str | None = None, alpha: float = 0.05, title: str | None = None, yname: str | None = None, ax=None)[source]#

Create a line plot of the prior, conventional, and posterior estimates.

Parameters:

  • column (ColumnType, optional) – Selected parameter. Defaults to None.

  • alpha (float, optional) – Sets the plot width. 0 is as wide as possible, 1 is as narrow as possible. Defaults to .05.

  • title (str, optional) – Plot title. Defaults to None.

  • yname (str, optional) – Name of the dependent variable. Defaults to None.

  • ax (AxesSubplot, optional) – Axis to write on.

Returns:

Plot.

Return type:

AxesSubplot

rank_conf_int(alpha: float = 0.05, columns: Sequence[int] | Sequence[str] | Sequence[bool] | None = None, simultaneous: bool = True) → ndarray[source]#

Compute rank confidence intervals.

Parameters:

  • alpha (float, optional) – Significance level. Defaults to 0.05.

  • columns (ColumnsType, optional) – Selected columns. Defaults to None.

  • simultaneous (bool, optional) – Indicates that rank confidence intervals should have correct simultaneous coverage. If False, the rank confidence intervals will have correct marginal coverage. Defaults to True.

Returns:

(# params, 2) array of rank confidence intervals.

Return type:

np.ndarray

rank_matrix_plot(title: str | None = None, **kwargs: Any)[source]#

Plot a heatmap of the rank matrix.

Parameters:

  • title (str, optional) – Plot title. Defaults to None.

  • **kwargs (Any) – Passed to sns.heatmap.

Returns:

Heatmap.

Return type:

AxesSubplot

rank_point_plot(yname: str | None = None, xname: Sequence[str] | None = None, title: str | None = None, columns: Sequence[int] | Sequence[str] | Sequence[bool] | None = None, ax=None, **kwargs: Any)[source]#

Create a point plot.

Parameters:

  • yname (str, optional) – Name of the endogenous variable. Defaults to None.

  • xname (Sequence[str], optional) – (# params,) sequence of parameter names. Defaults to None.

  • title (str, optional) – Plot title. Defaults to None.

  • columns (ColumnsType, optional) – Selected columns. Defaults to None.

  • ax (AxesSubplot, optional) – Axis to write on.

  • **kwargs (Any) – Passed to ResultsBase.conf_int().

Returns:

Plot.

Return type:

AxesSubplot

reconstruction_point_plot(yname: str | None = None, xname: Sequence[str] | None = None, title: str | None = None, alpha: float = 0.05, ax=None)[source]#

Create point plot of the reconstructed sample means.

Plots the distribution of sample means you would expect to see if this model were correct.

Parameters:

  • yname (str, optional) – Name of the endogenous variable. Defaults to None.

  • xname (Sequence[str], optional) – Names of x-ticks. Defaults to None.

  • title (str, optional) – Plot title. Defaults to None.

  • alpha – (float, optional): Plot the 1-alpha CI. Defaults to 0.05.

  • ax – (AxesSubplot, optional): Axis to write on.

Returns:

Plot.

Return type:

plt.axes._subplots.AxesSubplot