Multiple imputation of missing data under missing at random: including a collider as an auxiliary variable in the imputation model can induce bias

Will need to read this carefully

@dopapus Looks like the author list of this paper, which is generating consternation! https://t.co/zZIAjbGK5g

@PWGTennant @tomayates @stats_q Or so we all thought, until @camjpatrick went and read somewhere that including certain colliders in MI can transform the missingness mechanism from MAR to MNAR. https://t.co/UkIOAWFAkj

New Research: Multiple imputation of missing data under missing at random: including a collider as an auxiliary variable in the imputation model can induce bias #Epidemiology https://t.co/84Qv8jeVuJ