Balancing weights are used in observational causal inference to construct a population in which the treatment is independent of observed pretreatment covariates. The ideal weight is the ratio between the product of the covariate and treatment marginals and their joint distribution, for binary and continuous treatments alike. We propose Balancing Partition Tree (BPT) and Balancing Partition Forest (BPF), which estimate this ratio on a learned rectangular partition of covariate--treatment space. Each split maximizes the gain in a Bregman divergence computed from joint and marginal counts. Two choices of divergence yield criteria that maximize the mutual information captured by the partition or reduce the squared error of the weights. We show that the population residual of either criterion bounds the bias of inverse-probability-weighted and doubly robust response estimators that use the population partition weights. Across 17 synthetic and semi-synthetic settings and three response estimators, BPF obtains the lowest mean RMSE rank among seven weight estimators, and a single tree fits in 0.14 seconds on 16,512 observations.
Figures & tables
Figure 1: Balancing partition trees on 80 observed pairs. (a) Treatment–covariate dependence. (b) Each cell’s weight is computed from its joint and marginal counts. (c) The partition and its corresponding tree, with leaf weights shown numerically. Red cells are down-weighted; blue cells are up-weighted to approximate independence.
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 2: Fit-only wall-clock time on nested California subsamples with fixed selected configurations, on log–log axes. Solid lines use one thread; dashed lines allow up to four threads through each implementation’s existing parallelism. Points show the median of three repetitions; shading shows their min–max range. CBPS uses its exactly identified fallback for some fits.
Figure 3: MSM response curves of BPF, PW-GB and Unit, the constant weight w≡1 . Top: the four continuous synthetic settings at N=2,000 . Bottom: the three semi-synthetic settings. Lines show mean estimates over five seeds/folds, and shaded bands show the min–max range across them; the dashed black line is the true response μ(t) . Each semi-synthetic fold has its own response curve, so in these settings the dashed line shows the average of five true curves and the bands mostly reflect variation across those curves.
Figure 4: MSM dose-response curves, continuous synthetic settings, N=500 . Lines show mean estimates over five seeds/folds, and shaded bands show the min–max range across them; the dashed black line is the true response μ(t) .
Figure 5: MSM dose-response curves, continuous synthetic settings, N=2,000 . Lines show mean estimates over five seeds/folds, and shaded bands show the min–max range across them; the dashed black line is the true response μ(t) .
Figure 6: MSM dose-response curves, semi-synthetic settings. Lines show mean estimates over five seeds/folds, and shaded bands show the min–max range across them; the dashed black line is the true response μ(t) . Each semi-synthetic fold has its own response curve, so in these settings the dashed line shows the average of five true curves and the bands mostly reflect variation across those curves.
Probabilistic models are typically trained using task-agnostic objectives like log-loss, which can lead to significant errors in downstream estimation. This disconnect is especially critical in Inverse Probability Weighting (IPW) for causal inference, where propensity score errors near 0 and 1 often lead to high bias and variance. We propose a principled framework for deriving task-specific strictly proper scoring rules by matching the local curvature of the downstream error metric. We apply this to the Average Treatment Effect (ATE) estimation, deriving a closed-form loss and its corresponding canonical probability mapping that can be readily integrated with any model like a neural network or a gradient boosting algorithm. Extensive evaluations on causal inference benchmarks demonstrate that our tailored objective consistently outperforms standard likelihood-based and covariate-balancing approaches.
Roman Plaud, Alexandre Perez-Lebel, Antoine Saillenfest +4
Institut Polytechnique de Paris · Onepoint, France · Fundamental Technologies +2
Differential privacy is the leading mathematical framework for privacy protection, providing a probabilistic guarantee that safeguards individuals' private information when publishing statistics from a dataset. This guarantee is achieved by applying a randomized algorithm to the original data, which introduces unique challenges in data analysis by distorting inherent patterns. In particular, causal inference using observational data while preserving privacy is challenging because it requires a good covariate balance between treatment groups, but checking the covariate balance is challenging when privacy is a primary concern because releasing any statistic (e.g., t-statistic) for the balance check compromises privacy. Additionally, the performance of the privatized estimator critically depends on the choice of differential privacy mechanisms, and it remains unexplored how privacy-protecting causal inference could be conducted with observational data. In this article, we present a differentially private two-stage covariate balancing weighting estimator to infer causal effects from observational data. Our algorithm produces both point and interval estimators with statistical guarantees, such as covariate balance, rate optimality, and oracle budget allocation, under a given privacy budget.
Yuki Ohnishi, Jordan Awan
Department of Biostatistics, Yale School of Public Health · Department of Statistics, University of Pittsburgh
There is a precise sense in which drawing causal inferences from observational data is hard, even when identifiability is assumed. In particular, Robins and Ritov (1997) and Robins et al. (2003) showed that causal effects can be discontinuous as a function of the data distribution: two arbitrarily close data distributions might correspond to different causal effects. This is a fact independent of the choice of estimator; however, not all estimators are equally unstable. Our contribution is to surface a second layer of instability that depends on the choice of estimator. We show that many standard point estimates can be read as point summaries of multimodal distributions over the space of structural causal models. As such, estimators can jump discontinuously in the data distribution. This defines a taxonomy of estimators that admits a decision-theoretic reading: stability depends on whether the implicit loss function an estimator optimizes is aligned with the causal effect itself. Specifically, inverse propensity weighted estimators and regression estimators are examples of discontinuous summaries, while explicit posterior means and medians are shown to be continuous.