Probabilistic programming for embedding theory and quantifying uncertainty in econometric analysis
提出概率编程作为应对实证研究复制危机的方法，通过显式数据生成过程增强模型假设沟通、代码测试及理论与估计的一致性，并简化贝叶斯分析以改进不确定性建模与解释。
Abstract The replication crisis in empirical research calls for a more mindful approach to how we apply and report statistical models. For empirical research to have a lasting (policy) impact, these concerns are crucial. In this paper, we present Probabilistic Programming (PP) as a way forward. The PP workflow with an explicit data-generating process enhances the communication of model assumptions, code testing and consistency between theory and estimation. By simplifying Bayesian analysis, it also offers advantages for the interpretation, communication and modelling of uncertainty. We outline the advantages of PP to encourage its adoption in our community.