Struct Lplr 

pub struct Lplr { /* private fields */ }

Logistic partially linear regression for a binary outcome, P(Y = 1 | D, X) = expit(θ·D + r(X)), as DoubleML’s DoubleMLLPLR; θ is a log odds ratio.

Nuisances: ml_outcome (DoubleML’s ml_M) for P(Y = 1 | D, X), ml_t for E[logit M(D, X) | X], ml_m (and optionally ml_a) for E[D | X]. ml_M and ml_a are double cross-fitted (Lplr::n_folds_inner) and the score is solved with Newton’s method; see LplrScore for the scores.

Implementations

impl Lplr

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

ml_outcome learns M(D, X) = P(Y = 1 | D, X) (DoubleML’s ml_M), ml_t the log odds E[logit M(D, X) | X], ml_m the treatment regression E[D | X].

pub fn with_propensity( ml_outcome: impl Classifier + 'static, ml_t: impl Regressor + 'static, ml_m: impl Classifier + 'static, ) -> Self

As Lplr::new, with a classifier for a binary treatment’s propensity P(D = 1 | X).

pub fn from_boxed( ml_outcome: Box<dyn Classifier>, ml_t: Box<dyn Regressor>, ml_m: TreatmentLearner, ml_a: Option<TreatmentLearner>, ) -> Self

Boxed learners; ml_a defaults to a copy of ml_m.

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

pub fn ml_a(self, ml_a: TreatmentLearner) -> Self

A separate learner for a(X) = E[D | X], the regression used to build the preliminary estimate and r(X) (default: a copy of ml_m).

pub fn n_folds_inner(self, k: usize) -> Self

Inner folds of the double cross-fitting (default 5).

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

Number of outer cross-fitting folds (default 5).

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

Number of independent sample splits (default 1), aggregated by the median.

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

Seed for the outer and inner sample splits (default 0).

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

Draw seeded outer and inner sample splits and fit.

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

Fit on given outer splits (one Vec<Fold> per repetition); the inner splits are drawn from the seed.

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

Fit on given outer and inner splits: inner[r][k] are the inner folds (data indices, partitioning the training set of outer fold k) of repetition r.