Struct Ssm 

pub struct Ssm { /* private fields */ }

Average treatment effect of a binary treatment when the outcome is observed only where the selection indicator S is 1 (attrition, non-response), as DoubleML’s DoubleMLSSM. Requires a selection indicator (DmlData::with_selection_indicator) and, for SsmScore::Nonignorable, an instrument for selection (DmlData::with_instrument).

The score is doubly robust in the outcome models of the selected treated and controls, the propensity P(D = 1 | X) and the selection probability P(S = 1 | D, X). Under nonignorable selection each training fold is split in halves: the selection probability is learned on the first, and the other nuisances on the second with the predicted selection probability as a covariate.

Implementations

impl Ssm

pub fn new( ml_g: impl Regressor + 'static, ml_pi: impl Classifier + 'static, ml_m: impl Classifier + 'static, ) -> Self

ml_g learns E[Y | D, X] on the selected, ml_pi the selection probability P(S = 1 | D, X), ml_m the propensity P(D = 1 | X).

pub fn from_boxed( ml_g: Box<dyn Regressor>, ml_pi: Box<dyn Classifier>, ml_m: Box<dyn Classifier>, ) -> Self

pub fn score(self, score: SsmScore) -> Self

pub fn clipping_threshold(self, t: f64) -> Self

Treatment propensities are clipped to [t, 1 − t] (default t = 0.01). Selection probabilities are not clipped, as in DoubleML.

pub fn calibration(self, calibration: Calibration) -> Self

Calibrate the propensity scores before clipping (default: none), as DoubleML’s ps_processor_config.

pub fn normalize_ipw(self, normalize: bool) -> Self

Normalise the inverse-probability weights within treated and controls.

pub fn n_folds(self, n_folds: usize) -> Self

Number of cross-fitting folds (default 5).

pub fn n_rep(self, n_rep: usize) -> Self

Number of independent sample splits (default 1). Estimates are aggregated across repetitions by the median, as in DoubleML.

pub fn seed(self, seed: u64) -> Self

Seed for the sample splits (default 0).

pub fn fit(&self, data: &DmlData) -> Result<DmlFit>

Draw seeded sample splits and fit. With clustered data (DmlData::with_clusters) whole clusters are assigned to folds and standard errors are cluster-robust.

pub fn fit_with_cluster_folds( &self, data: &DmlData, splits: &[Vec<ClusterFold>], ) -> Result<DmlFit>

Fit clustered data with explicit cluster folds, one Vec<ClusterFold> per repetition.

pub fn sensitivity_benchmark( &self, data: &DmlData, long: &DmlFit, benchmarking_set: &[usize], ) -> Result<SensitivityBenchmark>

Sensitivity benchmark for the covariate columns benchmarking_set: refits this model without them (with the same sample splitting, so long must come from fit with the current settings) and compares it with long.

pub fn fit_with_folds( &self, data: &DmlData, splits: &[Vec<Fold>], ) -> Result<DmlFit>

Fit with caller-supplied splits, one Vec<Fold> per repetition (e.g. to reproduce another implementation exactly).

pub fn nuisance_regressor( self, name: &str, learner: impl Regressor + 'static, ) -> Self

Use learner for the regression nuisance name (DoubleML’s name, e.g. "ml_g1") instead of the model’s default learner for it.

pub fn nuisance_regressor_boxed( self, name: &str, learner: Box<dyn Regressor>, ) -> Self

As nuisance_regressor, boxed.

pub fn nuisance_classifier( self, name: &str, learner: impl Classifier + 'static, ) -> Self

Use learner for the classification nuisance name instead of the model’s default learner for it.

pub fn nuisance_classifier_boxed( self, name: &str, learner: Box<dyn Classifier>, ) -> Self

As nuisance_classifier, boxed.

Trait Implementations

impl LinearScoreModel for Ssm

fn strata(&self, data: &DmlData) -> Option<Vec<usize>>

Stratify by treatment and selection (D + 2S), as DoubleML.

fn name(&self) -> String

Label for summaries, e.g. "PLR (partialling out)".

fn validate(&self, data: &DmlData) -> Result<()>

Reject data the model cannot handle (called once before fitting).

fn estimate_score( &self, input: &ScoreInput<'_>, folds: &[Fold], ) -> Result<ScoreEstimate>

Cross-fit the nuisances on folds and return the score elements.