DAG-CLIP: A DAG Learning Framework in the Presence of Latent Variables
Abstract
Learning directed acyclic graphs (DAGs) to uncover causal mechanisms has attracted substantial attention in machine learning. While most existing methods focus exclusively on observed variables, many variables of substantive interest are latent constructs defined by statistical measurement models. Such constructs are particularly common in the social and behavioral sciences. In this paper, we propose a general statistical framework for DAG learning when some or all nodes of the graph are latent variables, accommodating non-Gaussian manifest variables such as binary and categorical data in the measurement model. To overcome the computational burden of high-dimensional integration in standard marginal likelihoods, we develop DAG learning via Composite-Likelihood-based screening and Iterative Pruning (DAG-CLIP), a two-step composite-likelihood-based learning algorithm. This algorithm first solves a smooth acyclicity-constrained optimization problem to screen out misspecified DAG structures, and subsequently performs BIC-guided composite-likelihood backward deletion within the Markov equivalence class (MEC) space to identify a sparse DAG. We establish the statistical consistency of the algorithm in recovering the MEC of the true DAG, and demonstrate its effectiveness through extensive simulation studies and a real-world application to large-scale educational survey data.
Related
- Data-Driven Learning of Directed Acyclic Graphs in Gaussian Linear Structural Equation Models with Latent Variables
- A Latent-Space Statistical Learning Framework for Semicontinuous Outcomes: Regularized Deviance and Boundary-Stabilized Optimization
- Wavelet-Enhanced CLIP with Noise Residual Learning for Generalizable Deepfake Detection
- Statistical Inference for Causal Discovery under Selection and Latent Variables via Single-Target Interventions
- Sequential reduction for discrete latent variables in ecological and evolutionary models using RTMB
BibTeX
@misc{zhang2026dagclip,
title = {{DAG-CLIP: A DAG Learning Framework in the Presence of Latent Variables}},
author = {Siliang Zhang and Yunxiao Chen and Irini Moustaki},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.06086},
eprint = {2610.06086},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.06086},
}
RIS
TY - UNPB
TI - DAG-CLIP: A DAG Learning Framework in the Presence of Latent Variables
AU - Siliang Zhang
AU - Yunxiao Chen
AU - Irini Moustaki
PY - 2026
DA - 2026/10/05
PB - arXiv
DO - 10.48550/arxiv.2610.06086
UR - https://arxiv.org/abs/2610.06086
AB - Learning directed acyclic graphs (DAGs) to uncover causal mechanisms has attracted substantial attention in machine learning. While most existing methods focus exclusively on observed variables, many variables of substantive interest are latent constructs defined by statistical measurement models. Such constructs are particularly common in the social and behavioral sciences. In this paper, we propose a general statistical framework for DAG learning when some or all nodes of the graph are latent variables, accommodating non-Gaussian manifest variables such as binary and categorical data in the measurement model. To overcome the computational burden of high-dimensional integration in standard marginal likelihoods, we develop DAG learning via Composite-Likelihood-based screening and Iterative Pruning (DAG-CLIP), a two-step composite-likelihood-based learning algorithm. This algorithm first solves a smooth acyclicity-constrained optimization problem to screen out misspecified DAG structures, and subsequently performs BIC-guided composite-likelihood backward deletion within the Markov equivalence class (MEC) space to identify a sparse DAG. We establish the statistical consistency of the algorithm in recovering the MEC of the true DAG, and demonstrate its effectiveness through extensive simulation studies and a real-world application to large-scale educational survey data.
ER -