Parameter Instability Sandbox

Explore how small dataset changes can flip the preferred p-adic model. This toy follows the chapter setup: model families y=x and y=2x, with a base set D and an augmented set D'.

Back to hub

Prime p

n (base range 1..n)

Loss mode

D includes points (i,i) for i=1..n and (i,2i) for i=1..n-1. D' adds (n,2n) and (n+1,2n+2).

Subset Instability Experiment

Re-sample random subsets and see how often each model wins under the chosen p-adic loss. This approximates sensitivity to finite observed data.

Subset size

Trials