A useful way to test any model is to push it to its extremes. In statistics, it is desirable for models to asymptotically converge to the true solution in the… Click to show full abstract
A useful way to test any model is to push it to its extremes. In statistics, it is desirable for models to asymptotically converge to the true solution in the limit of infinite sample size or zero noise. We have found that testing models with extreme examples is equally useful in geology. For example, Vermeesch (2012) used this approach to demonstrate that probability density plots break down when applied to large and/or high precision datasets. And Vermeesch (2018) used unrealistically large datasets to demonstrate that the youngest age peak is a poor estimator of the maximum depositional age because it drifts to younger ages with increasing sample size. Continuing in the same vein, Vermeesch and Tian (2014, 2018, hereafter referred to as VT1 and VT2) used extreme examples to highlight the fundamental differences between HeFTy and QTQt. The main thrust of the Comments by Gallagher and Ketcham (2017, 2019, hereafter referred to as GK1 and GK2) is that the case studies used by VT1 and VT2 are unrealistic. But this is exactly what they were meant to be.
               
Click one of the above tabs to view related content.