README- ::causal_graphCausalModel::causal_graph
- ::dataCausalModel::data
- ::do_operatorCausalModel::do_operator
- ::estimand_typeCausalModel::estimand_type
- ::estimate_effectCausalModel::estimate_effect
- ::get_common_causesCausalModel::get_common_causes
- ::get_effect_modifiersCausalModel::get_effect_modifiers
- ::get_estimatorCausalModel::get_estimator
- ::get_instrumentsCausalModel::get_instruments
- ::graph_with_rolesCausalModel::graph_with_roles
- ::identified_estimandCausalModel::identified_estimand
- ::identifies_variablesCausalModel::identifies_variables
- ::identify_effectCausalModel::identify_effect
- ::interpretCausalModel::interpret
- ::newCausalModel::new
- ::new_with_graph_stringCausalModel::new_with_graph_string
- ::proceeds_when_unidentifiableCausalModel::proceeds_when_unidentifiable
- ::refute_assess_overlapCausalModel::refute_assess_overlap
- ::refute_e_valueCausalModel::refute_e_value
- ::refute_estimateCausalModel::refute_estimate
- ::refute_graphCausalModel::refute_graph
- ::refute_linear_sensitivityCausalModel::refute_linear_sensitivity
- ::summaryCausalModel::summary
- ::unobserved_graph_variablesCausalModel::unobserved_graph_variables
- ::unused_data_variablesCausalModel::unused_data_variables
- ::view_modelCausalModel::view_model
- Vis_supported_graph_format
riantr/moonbit_dowhy/estimators does not have a README file
CausalEstimator
pub struct CausalEstimator {
target_estimand : IdentifiedEstimand
kind : EstimatorKind
weighting_scheme : WeightingScheme
min_ps_score : Double
max_ps_score : Double
num_strata : NumStrata
clipping_threshold : Int
instrument_name : String
rd_variable_name : String
rd_threshold_value : Double
rd_bandwidth : Double
rd_local_data : DataFrame?
model : OlsFit
first_stage : TwoStageModel?
second_stage : TwoStageModel?
glm_model : GlmFit
glm_family : GlmFamily
encoders : Encoders
features : Array[String]
propensities : Array[Double]
propensity_beta : Array[Double]
adjustment_names : Array[String]
effect_modifier_names : Array[String]
treatment_name : String
outcome_name : String
treatment_value : Double
control_value : Double
confidence_level : Double
num_simulations : Int
sample_size_fraction : Double
compute_confidence_intervals : Bool
test_significance : Bool
num_matches_per_unit : Int
distance_metric : DistanceMetric
}A causal estimator.
The estimator owns the identified estimand and the state that fit
produced; estimate_effect turns a data frame into a CausalEstimate.
CausalEstimator::adjustment_set
The adjustment variables the estimator uses.
CausalEstimator::base
The base estimator of the upstream.
This is the port of the base class CausalEstimator of the upstream, whose
placeholder methods raise a NotImplementedError. The port reports the
placeholders with the predicates supports_do_operator and
supports_symbolic_estimator and with a None return, because MoonBit
cannot catch the abort of a method.
The upstream builds its base estimator without an estimand. The port needs
one for every estimator, so the base takes the estimand it reports with.
CausalEstimator::control_value
The treatment value of the control arm.
The E-value analysis of the refuters package needs it, because the
contrast of the treatment scales the estimate of the upstream.
CausalEstimator::design_features
The columns of the design matrix of a regression based estimator, in order.
The first column is the column of ones of the intercept, the next one is the
treatment and the last ones are the adjustment variables and the effect
modifiers, the way build_design_matrix builds them.
CausalEstimator::distance_matching
fn CausalEstimator::distance_matching(estimand : IdentifiedEstimand, num_matches_per_unit? : Int, distance_metric? : DistanceMetric) -> CausalEstimatorA distance matching estimator for the given estimand.
The estimator matches every unit of the treated arm with the units of the
control arm whose common causes are the closest ones, and the effect is the
difference of the outcomes of the matched units.
The number of the matches per unit and the distance metric are the two
parameters of the upstream: the upstream takes the number of the matches
from its num_matches_per_unit and the metric from its
distance_metric.
CausalEstimator::distance_matching_error
The message of the error the upstream raises when the distance matching
estimator does not fit the data.
The upstream raises an exception when the treatment is not binary and when
the estimand has no common cause. MoonBit cannot catch the abort of a
function, so the port reports the two cases with this function, which
returns None when the estimator fits the data.
CausalEstimator::do_operator
The mean of the outcome when the treatment is set to the given value.
This is the port of dowhy.causal_estimator.CausalEstimator.do, which
returns the mean of the interventional outcomes of the data: the value of
the outcome under the intervention, and not the difference between the two
arms.
The upstream keeps the data it was fitted on and reads it when the caller
gives none. The port keeps no data in the estimator, so the data is an
argument.
The name of the upstream is do, and do is a reserved word of the
toolchain, so the port calls the method do_operator.
The base estimator of the upstream does not implement the do operator and
raises a NotImplementedError when it is called, so the port gives None
for it.
CausalEstimator::doubly_robust
fn CausalEstimator::doubly_robust(estimand : IdentifiedEstimand, min_ps_score : Double, max_ps_score : Double) -> CausalEstimatorA doubly robust estimator.
CausalEstimator::effect_modifiers
The effect modifiers the estimator uses.
The bootstrap refuter perturbs them, the way the upstream does.
CausalEstimator::estimand
The estimand the estimator was built for.
CausalEstimator::estimate_conditional_effects
fn CausalEstimator::estimate_conditional_effects(self : CausalEstimator, data : DataFrame, treatment_value : Double, control_value : Double) -> Array[(String, Double)]The conditional estimates of the effect, one per group of the effect
modifier.
The key of a group is the value of the discretised effect modifier, or the
values of the discretised effect modifiers when there is more than one.
The list is empty when the estimator has no effect modifier, which is what
the upstream reports as a None.
CausalEstimator::estimate_effect
fn CausalEstimator::estimate_effect(self : CausalEstimator, data : DataFrame, treatment_value : Double, control_value : Double, target_units : String) -> CausalEstimateEstimate the treatment effect.
target_units is one of "ate", "att" or "atc".
CausalEstimator::estimator_kind
The algorithm of the estimator.
CausalEstimator::fit
fn CausalEstimator::fit(self : CausalEstimator, data : DataFrame, effect_modifier_names : Array[String]) -> UnitFit the estimator on the given data.
The estimator remembers the adjustment variables of its estimand, fits the
model it needs and stores the propensity scores when it uses them.
CausalEstimator::generalized_linear_model
fn CausalEstimator::generalized_linear_model(estimand : IdentifiedEstimand, family : GlmFamily) -> CausalEstimatorA generalised linear model estimator for the given estimand.
The family decides the link of the model: a binomial family fits a logistic
regression, a poisson family fits a count model and the gaussian family
fits an ordinary least squares model.
CausalEstimator::glm_family
The family of the generalised linear model estimator.
The E-value analysis of the refuters package needs it, because the
conversion of the estimate to the risk ratio scale depends on the family.
CausalEstimator::glm_model
The fitted model of the generalised linear model estimator.
The E-value analysis of the refuters package needs it, because the
estimate of a generalised linear model is on the scale of the link of its
family.
CausalEstimator::instrument_name
The name of the instrument of an instrumental variable estimator.
It is empty for every other estimator.
CausalEstimator::instrumental_variable
fn CausalEstimator::instrumental_variable(estimand : IdentifiedEstimand, instrument_name : String) -> CausalEstimatorAn instrumental variable estimator.
CausalEstimator::intervene
The mean outcome of the intervention do(treatment = value).
This is the port of the do operator of the upstream estimators: the
treatment is set to the given value in every row of the data and the model
predicts the outcome of the modified data.
CausalEstimator::linear_regression
An ordinary least squares estimator for the given estimand.
CausalEstimator::outcome
The outcome column the estimator uses.
CausalEstimator::propensity_score_matching
A propensity score matching estimator.
CausalEstimator::propensity_score_stratification
fn CausalEstimator::propensity_score_stratification(estimand : IdentifiedEstimand, num_strata : NumStrata, clipping_threshold : Int) -> CausalEstimatorA propensity score stratification estimator.
CausalEstimator::propensity_score_weighting
fn CausalEstimator::propensity_score_weighting(estimand : IdentifiedEstimand, min_ps_score : Double, max_ps_score : Double, scheme : WeightingScheme) -> CausalEstimatorAn inverse propensity score weighting estimator.
CausalEstimator::propensity_scores
The propensity scores the estimator computed.
CausalEstimator::regression_discontinuity
fn CausalEstimator::regression_discontinuity(estimand : IdentifiedEstimand, rd_variable_name : String, rd_threshold_value : Double, rd_bandwidth : Double) -> CausalEstimatorA regression discontinuity estimator.
CausalEstimator::regression_model
The fitted model of a regression based estimator.
The sensitivity analysis of the refuters package needs the coefficients,
the standard errors, the t statistics and the degrees of freedom of the fit,
which are the numbers the upstream reads from the model of its
LinearRegressionEstimator.
CausalEstimator::set_compute_confidence_intervals
fn CausalEstimator::set_compute_confidence_intervals(self : CausalEstimator, compute : Bool) -> UnitAsk the estimator to compute the confidence intervals.
CausalEstimator::set_confidence_level
Set the confidence level of the intervals.
CausalEstimator::set_estimand
Replace the estimand the estimator was built for.
The refuters use this to estimate a different effect with the same
estimator.
CausalEstimator::set_num_simulations
Set the number of bootstrap simulations.
CausalEstimator::set_sample_size_fraction
Set the fraction of the sample a bootstrap resample takes.
CausalEstimator::set_test_significance
Ask the estimator to test the significance of the estimate.
CausalEstimator::strata_labels
The stratum of every row of the data.
This is the port of the strata column that the propensity score
stratification estimator of the upstream adds to the data. The upstream
searches for the number of the strata when its estimator holds the
automatic number, and the port searches for it the same way.
CausalEstimator::supports_do_operator
Whether the estimator has a do operator.
This is the port of the _do of the upstream. The base estimator of the
upstream does not implement it and raises a NotImplementedError when it
is called, and so do the estimators of the upstream that do not derive from
the regression estimator or from the doubly robust estimator, which are the
two classes that implement it.
The port computes the do operator for every estimator that has a model, so
this predicate tells what the upstream does and not what the port does.
CausalEstimator::supports_symbolic_estimator
Whether the estimator has a symbolic estimator.
This is the port of the construct_symbolic_estimator of the upstream. The
base estimator of the upstream does not implement it and raises a
NotImplementedError when it is called, and every estimator of the
upstream that the port covers implements it.
The port does not build the symbolic estimators, because they are the
symbolic expressions of the sympy of the upstream, so this predicate tells
what the upstream does and not what the port does.
CausalEstimator::symbolic_estimator
The symbolic estimator of the estimator.
The base estimator of the upstream does not build one and raises a
NotImplementedError when it is asked for, so the port gives None for
it, and so does the port for the instrumental variable estimator, whose
symbolic estimator of the upstream holds the expressions of the sympy.
The regression discontinuity estimator of the upstream builds the empty
string, so the port builds the empty string too.
The upstream builds the symbolic estimator when the estimator is fitted, and
the port computes it when it is asked for, so the effect modifiers of the
estimator are the ones of its last fit.
CausalEstimator::treatment
The treatment column the estimator uses.
CausalEstimator::treatment_value
The treatment value of the treated arm.
The E-value analysis of the refuters package needs it, because the
contrast of the treatment scales the estimate of the upstream.
CausalEstimator::two_stage_error
The error the two stage regression estimator reports for the estimand of
the estimator, or an empty string when it can estimate the effect.
The upstream raises a ValueError when the estimand holds no mediator
variable or more than one, and it raises an Exception when the estimand
holds more than one treatment variable or more than one outcome variable.
MoonBit cannot catch an error, so the port aborts with the message of the
upstream; this function reports the message of the error the upstream
raises, so that a caller can check for it before it fits the estimator.
CausalEstimator::two_stage_features
The columns of the design matrix of the two stages of the two stage
regression estimator.
The first element holds the columns of the first stage, the one that
regresses the mediator on the treatment, and the second one the columns of
the second stage, the one that regresses the outcome on the mediator.
CausalEstimator::two_stage_regression
A two stage regression estimator.
CausalEstimator::with_estimand
fn CausalEstimator::with_estimand(self : CausalEstimator, estimand : IdentifiedEstimand) -> CausalEstimatorA copy of the estimator that estimates a different estimand.
The refuters use this to re-estimate the effect with the same algorithm on
modified data. The copy starts without any fitted state, so the estimator
that was handed in keeps the model and the adjustment set it was fitted
with.
CausalEstimator::with_instrument_name
A copy of the estimator that uses a different instrument.
The placebo treatment refuter uses this for the instrumental variable
method, whose instrument the placebo replaces.
DistanceMetric
pub enum DistanceMetric {
Minkowski(Double)
Euclidean
Manhattan
Chebyshev
}The distance metric of the distance matching estimator.
The upstream hands the name of the metric to the NearestNeighbors of
scikit-learn, which uses the Minkowski metric with an exponent of two by
default. The port supports the Minkowski family, which is the family the
upstream uses with its default settings.
DistanceMetric::chebyshev
The Chebyshev distance.
DistanceMetric::euclidean
The Euclidean distance.
DistanceMetric::manhattan
The Manhattan distance.
DistanceMetric::minkowski
The Minkowski metric with the given exponent.
EstimatorKind
pub enum EstimatorKind {
Base
LinearRegression
PropensityScoreWeighting
PropensityScoreMatching
PropensityScoreStratification
DoublyRobust
InstrumentalVariable
TwoStageRegression
RegressionDiscontinuity
GeneralizedLinearModel
DistanceMatching
}The estimator family that DoWhy offers.
The kind decides which algorithm fit and estimate_effect run; the
estimator itself keeps the fitted state, exactly like the Python classes
inherit from CausalEstimator.
NumStrata
pub enum NumStrata {
Fixed(Int)
Auto
}The number of strata of the propensity score stratification estimator.
The upstream takes the string "auto", which searches for the number of
strata, and an integer, which is the number of strata.
NumStrata::auto
The number of strata is searched for, the way the upstream searches for it.
TwoStageModel
type TwoStageModelOne stage of the two stage regression estimator.
Every stage has its own encoders, because the treatment of the first stage
is the column the second stage regresses on.
WeightingScheme
pub enum WeightingScheme {
IpsWeight
IpsStabilizedWeight
IpsNormalizedWeight
}The weighting scheme of the propensity score weighting estimator.
WeightingScheme::ips_normalized_weight
The normalised inverse propensity score weighting scheme.
WeightingScheme::ips_stabilized_weight
The stabilised inverse propensity score weighting scheme.
WeightingScheme::ips_weight
The inverse propensity score weighting scheme.
WeightingScheme::name
The name of the weighting scheme.
DEFAULT_NUMBER_OF_SIMULATIONS_CI
let DEFAULT_NUMBER_OF_SIMULATIONS_CI : IntThe default number of simulations used for the confidence intervals.
DEFAULT_NUMBER_OF_SIMULATIONS_STAT_TEST
let DEFAULT_NUMBER_OF_SIMULATIONS_STAT_TEST : IntThe default number of simulations used for the significance test.
NUM_QUANTILES_TO_DISCRETIZE_CONT_COLS
let NUM_QUANTILES_TO_DISCRETIZE_CONT_COLS : IntThe default number of quantiles a continuous effect modifier is split into.
default_not_implemented_error_message
fn default_not_implemented_error_message() -> StringThe message of the error the upstream raises for a method it does not
implement.
This is the port of DEFAULT_NOTIMPLEMENTEDERROR_MSG of the upstream. The
upstream puts the class of the estimator in the message.
Install
Download zipMoonBit port of DoWhy 0.14 (PyWhy). Causal graph construction and analysis, effect identification (backdoor / frontdoor / complete adjustment), estimation (linear regression, generalized linear model, propensity score weighting / matching / stratification, doubly robust, instrumental variable, two stage, regression discontinuity), refutation (bootstrap, placebo treatment, data subset, dummy outcome, unobserved common cause, the assess overlap refuter that learns the rules of the support and of the overlap with boolean rules), intervention sampling (do samplers), temporal graphs, causal discovery (the PC algorithm with the Fisher-z test) and the portable core of the graphical causal models (the p-value merging, the bootstrap confidence intervals, the Kullback-Leibler divergences, the density estimators, the Shapley values, the independence tests, the prediction models, the model that selects itself out of a zoo of models, the causal mechanisms, the automatic assignment of the causal mechanisms with the selection of the model of a node by cross validation, the fitting of a causal model, the what-if queries (the interventional samples, the counterfactual samples and the average causal effect), the validation layer (the kernel based conditional independence test, the validation of a causal structure and the validation of an invertible model) and the falsification layer (the local Markov condition, the pairwise dependencies, the parental d-separations, the causal minimality and the permutation test of a graph) and the model evaluation layer (the normalised mean squared error, the continuous ranked probability score, the performance of the causal mechanisms, the test of the invertibility assumption and the report of the evaluation) and the distribution change layer (the test of the change of a mechanism, the attribution of the change of the distribution of a target node, the multiply-robust attribution of the change of the mean or of the variance of a target node and the scores of the change of the distribution of every node). Pure MoonBit with zero external runtime dependencies.
causal-inferencedowhycausal-graphcausal-discoverybackdoorfrontdoorpropensity-scoretreatment-effectrefutationoverlapidentifiersestimatorsshapley-valuesindependence-teststructural-causal-modelcausal-mechanismwhat-ifcounterfactual
MIT
Updated 1 hour ago
324 downloads