MedNexus
2025年 · 第63卷第12期
MedNexus
Selection bias means that the observation sample of the study is not randomly selected from the population, but is screened through some non-random mechanism. This screening mechanism makes the sample different from the population in certain characteristics, which leads to errors in sample-based statistical inference (including causal inference). The non-random sample selection mechanism causes the final analysis sample to not represent the target clinical population, which makes the effect estimation (such as treatment efficacy) biased. Selection graph is an important semantic extension to directed acyclic graph (DAG). By introducing a special node "S", it explicitly represents the sample selection mechanism, which is specifically used to deal with selection bias or sample selection problem.
本期目次

