There are two general views in causal analysis of experimental data: the super population view that the units are an independent sample from some hypothetical infinite populations, and the finite population view that the potential outcomes of the experimental units are fixed and the randomness comes solely from the physical randomization of the treatment assignment. These two views differs conceptually and mathematically, resulting in different sampling variances of the usual difference-in-means estimator of the average causal effect. Practically, however, these two views result in identical variance estimators. By recalling a variance decomposition and exploiting a completeness-type argument, we establish a connection between these two views in completely randomized experiments. This alternative formulation could serve as a template for bridging finite and super population causal inference in other scenarios.
Language Agents Meet Causality -- Bridging LLMs and Causal World Models
University of Amsterdam（阿姆斯特丹大学）；Institute for Logic, Language and Computation (ILLC)（逻辑、语言与计算研究所）；University of Edinburgh（爱丁堡大学）；Institute for Language, Cognition and Computation (ILCC)（语言、认知与计算研究所）
About forty years ago, in a now--seminal contribution, Rosenbaum & Rubin (1983) introduced a critical characterization of the propensity score as a central quantity for drawing causal inferences in observational study settings. In the decades since, much progress has been made across several research fronts in causal inference, notably including the re-weighting and matching paradigms. Focusing on the former and specifically on its intersection with machine learning and semiparametric efficiency theory, we re-examine the role of the propensity score in modern methodological developments. As Ro…
The health effects of environmental exposures have been studied for decades, typically using standard regression models to assess exposure-outcome associations found in observational non-experimental data. We propose and illustrate a different approach to examine causal effects of environmental exposures on health outcomes from observational data. Our strategy attempts to structure the observational data to approximate data from a hypothetical, but realistic, randomized experiment. This approach, based on insights from classical experimental design, involves four stages, and relies on modern c…
Meta-Router: Bridging Gold-standard and Preference-based Evaluations in Large Language Model Routing
Department of Statistics, Indiana University Bloomington（印第安纳大学布卢明顿分校统计系）；Department of Statistics, North Carolina State University（北卡罗来纳州立大学统计系）；Department of Biostatistics, The University of Texas MD Anderson Cancer Center（德克萨斯大学MD安德森癌症中心生物统计学系）
Instrumental variable methods are widely used in medical and social science research to draw causal conclusions when the treatment and outcome are confounded by unmeasured confounding variables. One important feature of such studies is that the instrumental variable is often applied at the cluster level, e.g., hospitals' or physicians' preference for a certain treatment where each hospital or physician naturally defines a cluster. This paper proposes to embed such observational instrumental variable data into a cluster-randomized encouragement experiment using statistical matching. Potential o…
Recent approaches in causal inference have proposed estimating average causal effects that are local to some subpopulation, often for reasons of efficiency. These inferential targets are sometimes data-adaptive, in that they are dependent on the empirical distribution of the data. In this short note, we show that if researchers are willing to adapt the inferential target on the basis of efficiency, then extraordinary gains in precision can be obtained. Specifically, when causal effects are heterogeneous, any asymptotically normal and root-$n$ consistent estimator of the population average caus…
Qualcomm AI Research（高通人工智能研究院）；University of Amsterdam（阿姆斯特丹大学）
Learning high-level causal representations together with a causal model from unstructured low-level data such as pixels is impossible from observational data alone. We prove under mild assumptions that this representation is however identifiable in a weakly supervised setting. This involves a dataset with paired samples before and after random, unknown interventions, but no further labels. We then introduce implicit latent causal models, variational autoencoders that represent causal variables and causal structure without having to optimize an explicit discrete graph structure. On simple image…
We propose to address the task of causal structure learning from data in a supervised manner. Existing work on learning causal directions by supervised learning is restricted to learning pairwise relation, and not well suited for whole DAG discovery. We propose a novel approach of modeling the whole DAG structure discovery as a supervised learning. To fit the problem in hand, we propose to use permutation equivariant models that align well with the problem domain. We evaluate the proposed approach extensively on synthetic graphs of size 10,20,50,100 and real data, and show promising results co…
Causal inference can estimate causal effects, but unless data are collected experimentally, statistical analyses must rely on pre-specified causal models. Causal discovery algorithms are empirical methods for constructing such causal models from data. Several asymptotically correct methods already exist, but they generally struggle on smaller samples. Moreover, most methods focus on very sparse causal models, which may not always be a realistic representation of real-life data generating mechanisms. Finally, while causal relationships suggested by the methods often hold true, their claims abou…
The Hong Kong University of Science and Technology（香港科技大学）；Zhejiang Laboratory（之江实验室）；Huawei Noah’s Ark Lab（华为诺亚方舟实验室）
This paper proposes a Disentangled gEnerative cAusal Representation (DEAR) learning method under appropriate supervised information. Unlike existing disentanglement methods that enforce independence of the latent variables, we consider the general case where the underlying factors of interests can be causally related. We show that previous methods with independent priors fail to disentangle causally related factors even under supervision. Motivated by this finding, we propose a new disentangled learning method called DEAR that enables causal controllable generation and causal representation le…
