Comment: Clarifying Endogeneous Data Structures and Consequent Modelling Choices Using Causal Graphs
Erica E. M. Moodie, David A. Stephens · Statistical Science · 2020
We read with great interest the article by Qian, Klasnja and Murphy (2020), and commend the authors for focusing on principled estimation and providing a quantitative approach to healthcare delivery through mobile devices.The quantitative analyses studied here could have wideranging applications that may serve to increase patient empowerment by taking medical monitoring and even intervention out of the clinic and into the home.Here, we wish to delve into two complementary aspects of the work: first, we attempt to give clarifications concerning the parameter(s) of interest, and second, we provide visualizations of potential scenarios that may help to clarify estimands and when biases due to endogeneity may arise. TREATMENT EFFECTS: ONE, TWOOR TOO MANY?