Network Effect Decomposition for Causal Effect Estimation under Network Interference
Abstract
Estimating causal effects from observational network data requires accounting for both individual treatment and neighborhood exposure. Existing outcome models commonly incorporate these intervention variables into a joint predictor and estimate causal effects through prediction contrasts, leaving the corresponding response components implicit. We propose the underlineNetwork underlineEffect underlineDecomposition underlineNetwork (NEDNet), which parameterizes the conditional outcome response as a query-invariant pre-treatment baseline plus separate treatment- and exposure-response pathways. Path-specific neighborhood contextualization supplies local pre-treatment information to the two pathways. We characterize the fitted response through a unique reference-normalized decomposition, which is equivalent to a zero mixed treatment--exposure contrast. Experiments on semi-synthetic benchmarks derived from BlogCatalog and Flickr show that NEDNet achieves the lowest PEHE among the compared methods for individual, peer, and total effects across homogeneous and covariate-heterogeneous response settings.
Related
- COSTA: Covariance-Optimized Design and Causal Inference under Network-Temporal Interference
- Unbiased Treatment Effect Estimation under Network Interference via Neighborhood-Excluded Cross-Fitting
- Spatial causal inference in the presence of unmeasured confounding and interference
- Interference-Free Capacity of a Binary Interference Channel with Causal Observation
- Causal Inference for Sequential Settings under Interference and Latent Confounding
BibTeX
@misc{huang2026network,
title = {{Network Effect Decomposition for Causal Effect Estimation under Network Interference}},
author = {Zhirong Huang and Wenzhen Zhang and Guangquan Lu and Boyan Chen and Debo Cheng and Shichao Zhang},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.07879},
eprint = {2611.07879},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.07879},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Network Effect Decomposition for Causal Effect Estimation under Network Interference
AU - Zhirong Huang
AU - Wenzhen Zhang
AU - Guangquan Lu
AU - Boyan Chen
AU - Debo Cheng
AU - Shichao Zhang
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.07879
UR - https://arxiv.org/abs/2611.07879
AB - Estimating causal effects from observational network data requires accounting for both individual treatment and neighborhood exposure. Existing outcome models commonly incorporate these intervention variables into a joint predictor and estimate causal effects through prediction contrasts, leaving the corresponding response components implicit. We propose the underlineNetwork underlineEffect underlineDecomposition underlineNetwork (NEDNet), which parameterizes the conditional outcome response as a query-invariant pre-treatment baseline plus separate treatment- and exposure-response pathways. Path-specific neighborhood contextualization supplies local pre-treatment information to the two pathways. We characterize the fitted response through a unique reference-normalized decomposition, which is equivalent to a zero mixed treatment--exposure contrast. Experiments on semi-synthetic benchmarks derived from BlogCatalog and Flickr show that NEDNet achieves the lowest PEHE among the compared methods for individual, peer, and total effects across homogeneous and covariate-heterogeneous response settings.
ER -