Weibull Shape Parameter Tests for Signal Detection

Implementation of Bayesian and frequentist Weibull Shape Parameter (WSP) tests for signal detection in pharmacovigilance based on right-censored time-to-event data to flag associations between drugs and adverse events. The WSP test is based on the assumption of constant hazard reflected by a Weibull type distribution with shape parameters equal to one. Based on the shape parameter estimates (posterior distribution or point estimate), the WSP test method performs a hypothesis test on each shape parameter and combines them to a decision on the presence of a signal. Methods described in Sauzet and Cornelius (2022) , Sauzet et al. (2024) , and Dyck and Sauzet (2025) .

WSPsignal

The family of Weibull Shape Parameter (WSP) tests was developed to detect signals of adverse events in electronic health records.

Most recently, the BPgWSP test was developed and presented in

Dyck, J., & Sauzet, O. (2025). The BPgWSP test: a Bayesian Weibull Shape Parameter signal detection test for adverse drug reactions. arXiv preprint arXiv:2412.05463.

available on https://arxiv.org/abs/2412.05463

triggering the effort to gather all WSP tests ready to apply in one package. A preprint describing the R package in more detail will be available soon.

Installation of the WSPsignal package

You can install the WSPsignal package with

remotes::install_github(repo = "julia-dyck/WSPsignal")

Tuning of the WSP test for your purpose

To tune the WSP test for your application, you can use the R scripts on https://osf.io/h7edn/overview as a template, and adjust sample scenarios as well as test alternatives accordingly.