Selection bias is often a challenging concept in epidemiology. One feature that distinguishes selection bias from confounding and measurement bias is, selection bias results from a change in the sample under study. It leads to a critical but, surprisingly, unsolved question. 1/
https://twitter.com/EpidemiologyLWW/status/1562411147546218496
Which sample/population does the selection bias refer to? The referent population before the selection process? Or the selected sample? 2/
To address this question and unify the various existing definitions in the literature, we propose a refined definition of selection bias by considering any bias away from the true causal effect in the referent population, due to selecting the sample, as selection bias. 3/
That is, selection bias is defined as the difference between the true causal effect in the referent population and the effect estimate in the selected sample. If we assume no confounding and measurement bias, then selection bias, on the additive scale, is 👇. 4/ 
Selection bias can be further classified into two types: type 1 selection bias due to restricting to one or more level(s) of a collider (or a descendant of a collider), and type 2 selection bias due to restricting to one or more level(s) of an effect measure modifier. 5/
Then we could rewrite our definition of selection bias into two parts 👇. Type 1 selection bias will result in a difference between the true causal effect in the selected sample and the effect estimate in the selected sample. 6/
Type 2 selection bias will result in a difference between the true causal effect in the referent population and the true causal effect in the selected sample. 7/
We also discussed how selection bias affects internal validity and external validity, and described methods to minimize/eliminate selection bias. A key takeaway is, as Sander Greenland mentioned in "the causal foundations of applied probability and statistics"👇 #epitwitter end/
• • •
Missing some Tweet in this thread? You can try to
force a refresh
