A new research poster from the School of Underlying Conditions uses a wholly synthetic database of reports to show how a single heavily reported entry hides the signals around it, and how a simple rule brings them back.
A simulation study from the School of Underlying Conditions tests a comparator that leaves the largest contributor out
The database at the centre of Slop University’s newest research poster has never received a report. Its 400 products and 250 event terms came out of a generator, and its 1.2 million entries describe nothing that happened to anyone. The authors regard this as the study’s chief convenience.
The Loud Neighbour: Masking in Disproportionality Analysis, and a Leave-One-Product-Out Background is the work of Dr Vigdis Emenike, Senior Lecturer, and Dr Brynjar Obiora, Senior Research Fellow, both of the School of Underlying Conditions. It concerns the screening methods by which large collections of reports are searched for patterns, and what becomes of the search when one entry is reported far more often than its neighbours. It draws no conclusion about any medicine.
A background is somebody else’s reading before it is yours. I was told that about water for ten years, and it was a pleasure to find the remark had somewhere else to go.
— Dr Vigdis Emenike, Senior Lecturer, School of Underlying Conditions
The poster aligns a synthetic evidence base, a planted-signal design and a comparator-adjustment rule within a single screening framework. Where one entry supplied 40 per cent of a category’s reports, the standard screen recovered 38 per cent of the signals planted there. The adjusted screen recovered 79 per cent.
The University counts the poster among the first returns on the School’s willingness to lend its habits of mind above ground. Dr Obiora said the School had been “glad to be of use somewhere it holds no lease”, and thanked the colleagues who “explained the vocabulary without once asking what we wanted it for”.
The full poster is available from the University’s research repository under an open licence, doi:10.5555/slop.v3a4df.