数学博士后科研流动站 - 博士后出站报告（Khelfaoui Issam博士）
报告题目：Applied Causal Machine Learning and Econometric Innovation for Health and Technological Systems
报 告 人：Khelfaoui Issam 博士（汕头大学）
报告时间：2026年1月24日 10:00
报告地点：公共卫生学院214会议室
Abstract: Health, time, and economic resources constitute the three fundamental constraints shaping human welfare and development. This postdoctoral research examines how technological and biological systems interact with these constraints by investigating the causal effects of information and communication technologies (ICTs) and data-driven health innovations on population health outcomes. Motivated by persistent global health inequalities and the rapid diffusion of digital infrastructure, the thesis advances beyond correlational analysis by integrating modern econometric identification strategies with causal machine learning (CML) frameworks to generate policy-relevant evidence. The empirical core of the thesis focuses on multi-country panel data from developing and low-income economies, primarily sourced from the World Bank and the Global Burden of Disease (GBD) project. Health outcomes are measured along complementary dimensions of time and burden, including life expectancy at birth (LEB), disability-adjusted life years (DALYs), and years of life lost (YLLs). Preliminary diagnostics reveal strong cross-sectional dependence, mixed orders of integration (I(0)/I(1)), and endogeneity between ICT adoption and health outcomes, necessitating advanced estimation strategies. To address these challenges, the research employs bias-corrected dynamic panel estimators, cross-sectionally augmented autoregressive distributed lag (CS-ARDL) models, spatial regression techniques, instrumental variable approaches, and Driscoll–Kraay robust inference. Empirical findings from the completed chapters demonstrate that modern digital ICTs—particularly mobile connectivity and internet penetration—exert statistically and economically significant improvements in population health, while legacy technologies such as fixed-line telephony display weak or adverse effects. These results are robust across alternative health indicators, spatial dependence structures, and identification strategies, highlighting the central role of digital inclusion in improving longevity and reducing premature mortality. Beyond these completed contributions, the thesis places causal machine learning at the core of its scientific agenda. Two published and under-review studies in microbiome research develop and formalize causal ML frameworks—including Double Machine Learning, directed acyclic graphs, and reproducibility-oriented reporting standards (STROBE-CML)—to bridge the gap between predictive performance and causal interpretability in health-related data. Building on this foundation, ongoing working papers extend the ICT–health nexus by incorporating causal random forests and hybrid econometric–ML workflows to capture non-linear effects, heterogeneous treatment responses, and counterfactual policy scenarios. Finally, the thesis documents statistical and econometric contributions to interdisciplinary collaborative research spanning nanotechnology, toxicology, biosensing, and agricultural economics, underscoring the transferability of causal and quantitative methods across scientific domains. Overall, this postdoctoral research establishes a coherent, scalable, and policy-oriented framework that unifies econometric rigor and causal machine learning to evaluate how technological and biological innovations shape health, time, and economic outcomes in developing systems.
