A recent paper by Shrier et al was recently sent to me: Challenges in interpreting results from ‘multiple regression’ when there is interaction between covariates. If you’re not familiar with conditional versus marginal estimates it’s a nice review of this concept and even includes a formula in the appendix allowing you to calculate the difference between the conditional and marginal estimates.

I read a lot of instrumental variable (IV) papers and most include a description of the IV assumptions. Given that IV analyses are popular in different fields, it isn’t surprising that there is a lot of variability in how they are stated. There is more than one way they can be stated accurately. But sometimes it’s clear there has been some confusion. Below is a causal graph I often encounter in the epidemiology literature:

We worry so much about measurement in epidemiology. (Or at least we should.) We teach about the bias and lack of precision that can occur due to measurement error. We ask participants in our studies to be as accurate as possible when filling out questionnaires to avoid measurement error to the extent possible. So where does this concern about measurement go when it comes to authorship?