Covariate balancing for high-dimensional samples in controlled experiments
In controlled experiments, achieving covariate balancing across all groups is crucial as it ensures that the estimated treatment effects are not confounded by the effects of covariates. This study proposes a mixed-integer nonlinear programming model to address the covariate balancing problem. Specif...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/182520 |
| _version_ | 1824457029899517952 |
|---|---|
| author | Luo, Xi Yan, Penggao Yan, Ran Wang, Shuaian |
| author2 | School of Civil and Environmental Engineering |
| author_facet | School of Civil and Environmental Engineering Luo, Xi Yan, Penggao Yan, Ran Wang, Shuaian |
| author_sort | Luo, Xi |
| collection | NTU |
| description | In controlled experiments, achieving covariate balancing across all groups is crucial as it ensures that the estimated treatment effects are not confounded by the effects of covariates. This study proposes a mixed-integer nonlinear programming model to address the covariate balancing problem. Specifically, we introduce a new covariate imbalance measure, which is the maximum discrepancy in both the first and second central moments between any two groups. The second central moment can effectively capture the correlation of covariates in a physical sense, which is crucial for partitioning high-dimensional samples. A mixed-integer nonlinear programming model is constructed to minimize the proposed measure to obtain the optimal partitioning results. The nonlinear model is then linearized to accelerate the optimization process. We conduct computational experiments based on simulated datasets, including one-dimensional, two-dimensional, and three-dimensional Gaussian distributed samples, and a real clinic trial dataset. Compared to the conventional discrepancy-based method, our method achieves a 54.81% and a 40.6% reduction in the maximum discrepancy of partitioning results in the two-dimensional simulated Gaussian samples and the real clinic trial dataset, respectively. These results demonstrate the superiority of the proposed model in partitioning high-dimensional samples with correlated covariates compared with the conventional discrepancy-based method. |
| first_indexed | 2025-02-19T04:03:30Z |
| format | Journal Article |
| id | ntu-10356/182520 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2025-02-19T04:03:30Z |
| publishDate | 2025 |
| record_format | dspace |
| spelling | ntu-10356/1825202025-02-05T08:00:10Z Covariate balancing for high-dimensional samples in controlled experiments Luo, Xi Yan, Penggao Yan, Ran Wang, Shuaian School of Civil and Environmental Engineering Engineering Covariate balance Partitioning problem In controlled experiments, achieving covariate balancing across all groups is crucial as it ensures that the estimated treatment effects are not confounded by the effects of covariates. This study proposes a mixed-integer nonlinear programming model to address the covariate balancing problem. Specifically, we introduce a new covariate imbalance measure, which is the maximum discrepancy in both the first and second central moments between any two groups. The second central moment can effectively capture the correlation of covariates in a physical sense, which is crucial for partitioning high-dimensional samples. A mixed-integer nonlinear programming model is constructed to minimize the proposed measure to obtain the optimal partitioning results. The nonlinear model is then linearized to accelerate the optimization process. We conduct computational experiments based on simulated datasets, including one-dimensional, two-dimensional, and three-dimensional Gaussian distributed samples, and a real clinic trial dataset. Compared to the conventional discrepancy-based method, our method achieves a 54.81% and a 40.6% reduction in the maximum discrepancy of partitioning results in the two-dimensional simulated Gaussian samples and the real clinic trial dataset, respectively. These results demonstrate the superiority of the proposed model in partitioning high-dimensional samples with correlated covariates compared with the conventional discrepancy-based method. 2025-02-05T08:00:10Z 2025-02-05T08:00:10Z 2024 Journal Article Luo, X., Yan, P., Yan, R. & Wang, S. (2024). Covariate balancing for high-dimensional samples in controlled experiments. Journal of the Operational Research Society, 2423362-. https://dx.doi.org/10.1080/01605682.2024.2423362 0160-5682 https://hdl.handle.net/10356/182520 10.1080/01605682.2024.2423362 2-s2.0-85208470061 2423362 en Journal of the Operational Research Society © 2024 The Operational Research Society. All rights reserved. |
| spellingShingle | Engineering Covariate balance Partitioning problem Luo, Xi Yan, Penggao Yan, Ran Wang, Shuaian Covariate balancing for high-dimensional samples in controlled experiments |
| title | Covariate balancing for high-dimensional samples in controlled experiments |
| title_full | Covariate balancing for high-dimensional samples in controlled experiments |
| title_fullStr | Covariate balancing for high-dimensional samples in controlled experiments |
| title_full_unstemmed | Covariate balancing for high-dimensional samples in controlled experiments |
| title_short | Covariate balancing for high-dimensional samples in controlled experiments |
| title_sort | covariate balancing for high dimensional samples in controlled experiments |
| topic | Engineering Covariate balance Partitioning problem |
| url | https://hdl.handle.net/10356/182520 |
| work_keys_str_mv | AT luoxi covariatebalancingforhighdimensionalsamplesincontrolledexperiments AT yanpenggao covariatebalancingforhighdimensionalsamplesincontrolledexperiments AT yanran covariatebalancingforhighdimensionalsamplesincontrolledexperiments AT wangshuaian covariatebalancingforhighdimensionalsamplesincontrolledexperiments |