Questioning Bayesian Inference‐Based Models for Estimating Culprit Drugs for Adverse Drug Reactions

Reply to “Questioning Bayesian Inference‐Based Models for Estimating Culprit Drugs for Adverse Drug Reactions”

“…Noguchi et al 1 . noted room for improvement in the evaluation approach for the bayesian method proposed in ref.…”

“…While precision‐recall curves ( Figure ) were provided as suggested in ref. 1 solely for comparing relative effectiveness among scores, it is worth noting that the expected baseline probability of precision 4 varies depending on the number of prescribed drugs ( N ) for the ranking task; the chance level of being a culprit drug decreases as N increases. Interpreting precision‐recall curves compared with receiver operating characteristic curves requires a higher caution in our case.…”

“…Noguchi et al 1 noted room for improvement in the evaluation approach for the bayesian method proposed in ref. 2 to estimate causative drugs related to adverse events.…”