因果推断是通过研究处理效应发生的条件从而得到因果关系结论的过程。在实际的传染性疾病干预政策中，评估不同干预政策在不同地区的传染性疾病疫情发展效果的差异性，也就是判断处理效应的差异性，是因果推断在政策研究分析中的重要运用。然而，干预政策在同一地区实施前后、不同干预政策在不同地区实施、相同干预政策在不同地区实施的处理效应的差异性，为因果推断研究带来巨大的挑战。本项目以COVID-19疫情实施的干预政策为契机，采用统计方法推断COVID-19传播过程中的干预政策的因果关系。在因果效应的建模、识别和评估的总体思路中，对干预政策的传染性疾病的发展进行动态传播模型的预估，对干预政策因果推断模型中的双重差分和合成控制法中的参数进行改进，对干预政策的异质性处理效应构建新的统计量，利用深度学习进行参数估计，并对异质性处理效应分布进行检验，从而提高因果效应的评估效果，为服务于长远的政策提供理论支撑和实践基础。

Cause inference is the process of obtaining conclusions on causal connections based on the conditions of the occurrence of effects. For the intervention policies of infectious diseases, how to evaluate the differences from the spread of epidemic situations with a series of intervention policies in various hit areas, that is, to examine the differences of treatment effects, is an important application of causal inference in policy research and analysis. However, the different treatment effects of intervention policies before and after implementation in the same region, different intervention policies in different regions, and the implementation of the same intervention policy in different regions bring great challenges to the study of causal inference. This project takes the intervention policies of COVID-19 as an example, employing statistical inference to investigate the causality of interventions in the transmission of COVID-19. As a key idea of modeling, identifying and evaluating the causal effects of interventions, we estimate the spread of infectious diseases with interventions by dynamic transmission models, establish the improvement of the causal inference model parameters in difference-in-differences and synthetic control methods and construct the new statistics of the heterogeneous causal effect by deep learning. In doing so, accurate and precise estimates of causal effects could be obtained to provide theoretical support and practical basis for long-term policies.