Joint evaluation of accuracy parameters in diagnostic studies
Final Report Abstract
The general aim of the project was to improve the toolbox for analysing diagnostic studies, overcoming the current practice of separate evaluation of sensitivity and specificity by allowing for a clinically meaningful compensation. This aim could be reached by fully developing the concept of “comparison regions“. This is a general approach to visualize the uncertainty of a two-dimensional parameter estimate, supplementing the existing concept of two-dimensional confidence regions. Whereas confidence regions support the post-hoc testing of single-point hypotheses, comparison regions support the post-hoc application of one-sided tests on linear combinations of the two parameters – or even more generally of tests aiming at demonstrating that the true parameter values are within a convex region of interest. This is exactly what is needed in diagnostic research, as we are interested to demonstrate that sensitivity and specificity (or another pair of diagnostic accuracy parameters) are together of sufficient magnitude.
Publications
- On the use of comparison regions in visualizing stochastic uncertainty in some two-parameter estimation problems. Biometrical Journal. 2020; 62: 598-609
Eckert M, Vach W.
(See online at https://doi.org/10.1002/bimj.201800232)
