CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: The variety of no direct effect assumptions in dynamic treatment regimes Authors:  Lin Liu - Shanghai Jiao Tong University (China) [presenting]
Abstract: Recent theoretical results on semiparametric efficiency theory when some direct effects are known to be absent based on background knowledge are presented. Then we will showcase how such results can be applied to precision medicine or personalized decision making to more efficiently utilize the data. In particular, we will compare and contrast several works that leverage such ``no direct effect'' assumptions, and show that we can provide a unified picture of all these works. If time permitted, we will also present our adventure on how neural networks can be used to estimate parameters efficiently in this context, i.e. under the ``no direct effect'' assumption.