RESPONSE TO INVITED COMMENTARY: Shahar Responds to “Causal Diagrams and Measurement Bias”

摘要

Hernán and Cole (1) have written an article about causal diagrams and measurement (information) bias, a topic to which I contributed a gentle introduction (2). Their article follows the presentation of an abstract by Hernán and Robins (3), but the application to body mass index (BMI; weight (kg)/height (m)2) is, in my view, erroneous. Moreover, the proposal to display unspecified “constructs” as nodes in causal diagrams should be rejected.nnReferring to Figure 3 in their article, Hernán and Cole write, “This causal diagram is essentially equivalent to that proposed by Shahar… except that we allow for dependent measurement errors UWH for weight and height” (1, p. 957). To help the reader judge the truth of their statement, I redrew my diagram in the notation and layout of theirs; Figure 1 shows the 2 diagrams side by side.