Struct PropensityScoreMatching 

pub struct PropensityScoreMatching {
    pub weights: Vec<f64>,
    pub n_covariates: usize,
}

Propensity score matching for observational causal inference.

Estimates the probability of treatment assignment P(T=1|X) using logistic regression, then matches treated and control units.

Fields

weights: Vec<f64>

Logistic regression weights (length = n_covariates + 1, with intercept).

n_covariates: usize

Number of covariates.

Implementations

impl PropensityScoreMatching

pub fn new(n_covariates: usize) -> Self

Create a new propensity score matcher.

Arguments

• n_covariates — number of covariate dimensions

pub fn fit( &mut self, covariates: &[Vec<f64>], treatment: &[f64], lr: f64, n_iter: usize, )

Fit logistic regression to estimate propensity scores via gradient descent.

Arguments

• covariates — matrix of covariates, shape [n_obs][n_covariates]
• treatment — binary treatment indicator (0 or 1), length n_obs
• lr — learning rate
• n_iter — number of gradient descent iterations

pub fn predict_one(&self, x: &[f64]) -> f64

Predict propensity score P(T=1|X=x) for a single observation.

pub fn predict(&self, covariates: &[Vec<f64>]) -> Vec<f64>

Predict propensity scores for all observations.

pub fn estimate_ate( &self, covariates: &[Vec<f64>], treatment: &[f64], outcome: &[f64], ) -> f64

Estimate the Average Treatment Effect (ATE) using IPW (Inverse Probability Weighting).

ATE = E[Y(1)] - E[Y(0)] = E[T*Y/e(X)] - E[(1-T)*Y/(1-e(X))]

pub fn estimate_att( &self, covariates: &[Vec<f64>], treatment: &[f64], outcome: &[f64], ) -> f64

Estimate the Average Treatment Effect on the Treated (ATT).

ATT = E[Y(1)-Y(0)|T=1]

Trait Implementations

impl Clone for PropensityScoreMatching

fn clone(&self) -> PropensityScoreMatching

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PropensityScoreMatching

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.