A predictive model makes outcome predictions based on some given features, i.e., it estimates the conditional probability of the outcome given a feature vector. In general, a predictive model cannot estimate the causal effect of a feature on the outcome, i.e., how the outcome will change if the feature is changed while keeping the values of other features unchanged. This is because causal effect estimation requires interventional probabilities. However, many real world problems such as personalised decision making, recommendation, and fairness computing, need to know the causal effect of any feature on the outcome for a given instance. This is different from the traditional causal effect estimation problem with a fixed treatment variable. This paper first tackles the challenge of estimating the causal effect of any feature (as the treatment) on the outcome w.r.t. a given instance. The theoretical results naturally link a predictive model to causal effect estimations and imply that a predictive model is causally interpretable when the conditions identified in the paper are satisfied. The paper also reveals the robust property of a causally interpretable model. We use experiments to demonstrate that various types of predictive models, when satisfying the conditions identified in this paper, can estimate the causal effects of features as accurately as state-of-the-art causal effect estimation methods. We also show the potential of such causally interpretable predictive models for robust predictions and personalised decision making.
相關內容
In this paper, we design sub-linear space streaming algorithms for estimating three fundamental parameters -- maximum independent set, minimum dominating set and maximum matching -- on sparse graph classes, i.e., graphs which satisfy $m=O(n)$ where $m,n$ is the number of edges, vertices respectively. Each of the three graph parameters we consider can have size $\Omega(n)$ even on sparse graph classes, and hence for sublinear-space algorithms we are restricted to parameter estimation instead of attempting to find a solution.
Existing heterogeneous treatment effects learners, also known as conditional average treatment effects (CATE) learners, lack a general mechanism for end-to-end inter-treatment information sharing, and data have to be split among potential outcome functions to train CATE learners which can lead to biased estimates with limited observational datasets. To address this issue, we propose a novel deep learning-based framework to train CATE learners that facilitates dynamic end-to-end information sharing among treatment groups. The framework is based on \textit{soft weight sharing} of \textit{hypernetworks}, which offers advantages such as parameter efficiency, faster training, and improved results. The proposed framework complements existing CATE learners and introduces a new class of uncertainty-aware CATE learners that we refer to as \textit{HyperCATE}. We develop HyperCATE versions of commonly used CATE learners and evaluate them on IHDP, ACIC-2016, and Twins benchmarks. Our experimental results show that the proposed framework improves the CATE estimation error via counterfactual inference, with increasing effectiveness for smaller datasets.
Outcome phenotype measurement error is rarely corrected in comparative effect estimation studies in observational pharmacoepidemiology. Quantitative bias analysis (QBA) is a misclassification correction method that algebraically adjusts person counts in exposure-outcome contingency tables to reflect the magnitude of misclassification. The extent QBA minimizes bias is unclear because few systematic evaluations have been reported. We empirically evaluated QBA impact on odds ratios (OR) in several comparative effect estimation scenarios. We estimated non-differential and differential phenotype errors with internal validation studies using a probabilistic reference. Further, we synthesized an analytic space defined by outcome incidence, uncorrected ORs, and phenotype errors to identify which combinations produce invalid results indicative of input errors. We evaluated impact with relative bias [(OR-ORQBA)]/OR*100%]. Results were considered invalid if any contingency table cell was corrected to a negative number. Empirical bias correction was greatest in lower incidence scenarios where uncorrected ORs were larger. Similarly, synthetic bias correction was greater in lower incidence settings with larger uncorrected estimates. The invalid proportion of synthetic scenarios increased as uncorrected estimates increased. Results were invalid in common, low incidence scenarios indicating problematic inputs. This demonstrates the importance of accurately and precisely estimating phenotype errors before implementing QBA in comparative effect estimation studies.
Simulation-based inference (SBI) enables amortized Bayesian inference for simulators with implicit likelihoods. But when we are primarily interested in the quality of predictive simulations, or when the model cannot exactly reproduce the observed data (i.e., is misspecified), targeting the Bayesian posterior may be overly restrictive. Generalized Bayesian Inference (GBI) aims to robustify inference for (misspecified) simulator models, replacing the likelihood-function with a cost function that evaluates the goodness of parameters relative to data. However, GBI methods generally require running multiple simulations to estimate the cost function at each parameter value during inference, making the approach computationally infeasible for even moderately complex simulators. Here, we propose amortized cost estimation (ACE) for GBI to address this challenge: We train a neural network to approximate the cost function, which we define as the expected distance between simulations produced by a parameter and observed data. The trained network can then be used with MCMC to infer GBI posteriors for any observation without running additional simulations. We show that, on several benchmark tasks, ACE accurately predicts cost and provides predictive simulations that are closer to synthetic observations than other SBI methods, especially for misspecified simulators. Finally, we apply ACE to infer parameters of the Hodgkin-Huxley model given real intracellular recordings from the Allen Cell Types Database. ACE identifies better data-matching parameters while being an order of magnitude more simulation-efficient than a standard SBI method. In summary, ACE combines the strengths of SBI methods and GBI to perform robust and simulation-amortized inference for scientific simulators.
Treatment effect estimation from observational data is a central problem in causal inference. Methods based on potential outcomes framework solve this problem by exploiting inductive biases and heuristics from causal inference. Each of these methods addresses a specific aspect of treatment effect estimation, such as controlling propensity score, enforcing randomization, etc., by designing neural network architectures and regularizers. In this paper, we propose an adaptive method called Neurosymbolic Treatment Effect Estimator (NESTER), a generalized method for treatment effect estimation. NESTER brings together the ideas used in existing methods based on multi-head neural networks for treatment effect estimation into one framework. To perform program synthesis, we design a Domain Specific Language (DSL) for treatment effect estimation based on inductive biases used in literature. We also theoretically study NESTER's capability for treatment effect estimation. Our comprehensive empirical results show that NESTER performs better than state-of-the-art methods on benchmark datasets without compromising run time requirements.
To check the accuracy of Bayesian computations, it is common to use rank-based simulation-based calibration (SBC). However, SBC has drawbacks: The test statistic is somewhat ad-hoc, interactions are difficult to examine, multiple testing is a challenge, and the resulting p-value is not a divergence metric. We propose to replace the marginal rank test with a flexible classification approach that learns test statistics from data. This measure typically has a higher statistical power than the SBC rank test and returns an interpretable divergence measure of miscalibration, computed from classification accuracy. This approach can be used with different data generating processes to address likelihood-free inference or traditional inference methods like Markov chain Monte Carlo or variational inference. We illustrate an automated implementation using neural networks and statistically-inspired features, and validate the method with numerical and real data experiments.
Recent critiques of Physics Education Research (PER) studies have revoiced the critical issues when drawing causal inferences from observational data where no intervention is present. In response to a call for a "causal reasoning primer", this paper discusses some of the fundamental issues underlying statistical causal inference. In reviewing these issues, we discuss well-established causal inference methods commonly applied in other fields and discuss their application to PER. Using simulated data sets, we illustrate (i) why analysis for causal inference should control for confounders but not control for mediators and colliders and (ii) that multiple proposed causal models can fit a highly correlated data set. Finally, we discuss how these causal inference methods can be used to represent and explain existing issues in quantitative PER. Throughout, we discuss a central issue: quantitative results from observational studies cannot support a researcher's proposed causal model over other alternative models. To address this issue, we propose an explicit role for observational studies in PER that draw statistical causal inferences: proposing future intervention studies and predicting their outcomes. Mirroring a broader connection between theoretical motivating experiments in physics, observational studies in PER can make quantitative predictions of the causal effects of interventions, and future intervention studies can test those predictions directly.
Causality can be described in terms of a structural causal model (SCM) that carries information on the variables of interest and their mechanistic relations. For most processes of interest the underlying SCM will only be partially observable, thus causal inference tries to leverage any exposed information. Graph neural networks (GNN) as universal approximators on structured input pose a viable candidate for causal learning, suggesting a tighter integration with SCM. To this effect we present a theoretical analysis from first principles that establishes a novel connection between GNN and SCM while providing an extended view on general neural-causal models. We then establish a new model class for GNN-based causal inference that is necessary and sufficient for causal effect identification. Our empirical illustration on simulations and standard benchmarks validate our theoretical proofs.
A fundamental goal of scientific research is to learn about causal relationships. However, despite its critical role in the life and social sciences, causality has not had the same importance in Natural Language Processing (NLP), which has traditionally placed more emphasis on predictive tasks. This distinction is beginning to fade, with an emerging area of interdisciplinary research at the convergence of causal inference and language processing. Still, research on causality in NLP remains scattered across domains without unified definitions, benchmark datasets and clear articulations of the remaining challenges. In this survey, we consolidate research across academic areas and situate it in the broader NLP landscape. We introduce the statistical challenge of estimating causal effects, encompassing settings where text is used as an outcome, treatment, or as a means to address confounding. In addition, we explore potential uses of causal inference to improve the performance, robustness, fairness, and interpretability of NLP models. We thus provide a unified overview of causal inference for the computational linguistics community.
Recent contrastive representation learning methods rely on estimating mutual information (MI) between multiple views of an underlying context. E.g., we can derive multiple views of a given image by applying data augmentation, or we can split a sequence into views comprising the past and future of some step in the sequence. Contrastive lower bounds on MI are easy to optimize, but have a strong underestimation bias when estimating large amounts of MI. We propose decomposing the full MI estimation problem into a sum of smaller estimation problems by splitting one of the views into progressively more informed subviews and by applying the chain rule on MI between the decomposed views. This expression contains a sum of unconditional and conditional MI terms, each measuring modest chunks of the total MI, which facilitates approximation via contrastive bounds. To maximize the sum, we formulate a contrastive lower bound on the conditional MI which can be approximated efficiently. We refer to our general approach as Decomposed Estimation of Mutual Information (DEMI). We show that DEMI can capture a larger amount of MI than standard non-decomposed contrastive bounds in a synthetic setting, and learns better representations in a vision domain and for dialogue generation.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191