Loading...
|
Please use this identifier to cite or link to this item:
https://nccur.lib.nccu.edu.tw/handle/140.119/158569
|
| Title: | 跨期隨機實驗下的異質處方效果校準
Calibration of Heterogeneous Treatment Effects in Longitudinal Randomized Experiments |
| Authors: | 蘇娜玉
Su, Na-Yu |
| Contributors: | 莊皓鈞
周彥君
Chuang, Hao-Chun
Chou, Yen-Chun
蘇娜玉
Su, Na-Yu |
| Keywords: | 異質處方效果
線性校準
非線性誤差
交互式固定效應
高斯過程迴歸
數位行銷
Heterogeneous Treatment Effects
Linear Calibration
Nonlinear Bias
Interactive Fixed Effects
Gaussian Process Regression
Digital Marketing |
| Date: | 2025 |
| Issue Date: | 2025-08-04 14:25:49 (UTC+8) |
| Abstract: | 異質處方效果（Heterogeneous Treatment Effects, HTEs）估計在個人化數位行銷決策中扮演關鍵角色，現有機器學習方法的預測結果常因校準不良（uncalibrated）導致資源錯誤配置。Leng和Dimmery（2024）對此提出線性校準架構調整未校準預測值，提升其與真實因果效果的一致性。本研究基於台灣某金融機構六輪A/B隨機實驗資料，探討線性校準於跨期隨機實驗情境之適用性與限制。實證結果顯示，儘管標準化與正規化能降低部分誤差，線性校準仍無法顯著改善預測偏差與排名一致性，反映單純線性假設難以捕捉效果隨時間演變的非線性誤差結構。為此，本文進一步提出結合高斯過程迴歸（Gaussian Process Regression, GPR）的加法校準模型，以模擬最大概似估計（Simulated Maximum Likelihood Estimation, SMLE）同時擬合線性與非線性誤差，試圖提升校準模型表現。結果顯示，加法校準模型於此實證資料中仍受高雜訊限制，未能明顯提升校準成效。整體而言，本研究探討多輪異質性處方效果模型校準的挑戰，並提出結合非線性調整的方向，為未來優化校準方法提供理論與實務參考。
Heterogeneous Treatment Effects (HTEs) estimation is crucial for personalized decision-making in digital marketing. However, predictions from machine learning models are often uncalibrated, leading to inefficient resource allocation. To address this, Leng and Dimmery (2024) proposed a linear calibration method to better align predictions with true causal effects. This study evaluates its applicability using 6 rounds of A/B experiments from a financial institution in Taiwan. Results show that while standardization and normalization reduce some errors, linear calibration does not significantly improve bias or ranking consistency, suggesting that linear assumptions fail to capture the evolving nonlinear bia structure over time. To improve calibration, we propose an additive model combining Gaussian Process Regression (GPR) and Simulated Maximum Likelihood Estimation (SMLE) to correct both linear and nonlinear bias. However, high noise levels in the data still limit its effectiveness. This study highlights key challenges in calibrating HTEs models across multiple experimental rounds and suggests a direction for integrating nonlinear adjustments in future methods. |
| Reference: | Athey, S., & Imbens, G. (2016). Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of Sciences, 113(27), 7353-7360.
Bai, J. (2009). Panel data models with interactive fixed effects. Econometrica, 77(4), 1229-1279.
Finkelstein, A., & Hendren, N. (2020). Welfare analysis meets causal inference. Journal of Economic Perspectives, 34(4), 146-167.
Gordon, B. R., Zettelmeyer, F., Bhargava, N., & Chapsky, D. (2019). A comparison of approaches to advertising measurement: Evidence from big field experiments at Facebook. Marketing Science, 38(2), 193-225.
Hahn, P. R., Carvalho, C. M., Puelz, D., & He, J. (2018). Regularization and confounding in linear regression for treatment effect estimation. Bayesian Analysis, 13(1), 163–182.
Hill, J. L. (2011). Bayesian nonparametric modeling for causal inference. Journal of Computational and Graphical Statistics, 20(1), 217-240.
Kennedy, E. H. (2023). Towards optimal doubly robust estimation of heterogeneous causal effects. Electronic Journal of Statistics, 17(2), 3008-3049.
Kochenderfer, M. J., & Wheeler, T. A. (2019). Algorithms for optimization. Cambridge, MA: MIT Press.
Künzel, S. R., Sekhon, J. S., Bickel, P. J., & Yu, B. (2019). Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the National Academy of Sciences, 116(10), 4156-4165.
Lada, A., Peysakhovich, A., Aparicio, D., & Bailey, M. (2019, June). Observational data for heterogeneous treatment effects with application to recommender systems. In Proceedings of the 2019 ACM Conference on Economics and Computation (pp. 199-213).
Leng, Y., & Dimmery, D. (2024). Calibration of heterogeneous treatment effects in randomized experiments. Information Systems Research, 35(4), 1721-1742.
Nie, X., & Wager, S. (2021). Quasi-oracle estimation of heterogeneous treatment effects. Biometrika, 108(2), 299-319.
Prosperi, M., Guo, Y., Sperrin, M., Koopman, J. S., Min, J. S., He, X., ... & Bian, J. (2020). Causal inference and counterfactual prediction in machine learning for actionable healthcare. Nature Machine Intelligence, 2(7), 369-375.
Robinson, P. M. (1988). Root-N-consistent semiparametric regression. Econometrica: Journal of the Econometric Society, 931-954.
Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41-55.
Syrgkanis, V., Lei, V., Oprescu, M., Hei, M., Battocchi, K., & Lewis, G. (2019). Machine learning estimation of heterogeneous treatment effects with instruments. Advances in Neural Information Processing Systems, 32.
Wager, S., & Athey, S. (2018). Estimation and inference of heterogeneous treatment effects using random forests. Journal of the American Statistical Association, 113(523), 1228-1242.
Zhou, H., Li, S., Jiang, G., Zheng, J., & Wang, D. (2023, June). Direct heterogeneous causal learning for resource allocation problems in marketing. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 37, No. 4, pp. 5446-5454). |
| Description: | 碩士
國立政治大學
資訊管理學系
112356010 |
| Source URI: | http://thesis.lib.nccu.edu.tw/record/#G0112356010 |
| Data Type: | thesis |
| Appears in Collections: | [資訊管理學系] 學位論文
|
Files in This Item:
| File |
Description |
Size | Format | |
|---|
| 601001.pdf | | 1680Kb | Adobe PDF | 0 | View/Open |
|
All items in 政大典藏 are protected by copyright, with all rights reserved.
|
