Speaker
Description
In this talk I'll explore how we can use Shannon entropy and the Surprise to
quantify discordances between datasets. The Surprise is a tool based on the
Kullback-Leibler divergence and offers a way to quantify discordance between
datasets in multiple dimensions in parameter space. I’ll analyze Supernovae, time
delay gravitational lensing, BAO and CMB data for LambdaCDM model and
variations with one more parameter. We’ll compare both measures of discordance
with the usual measure of distances between marginalized posterior distributions
and see how they relate to each other in the context of the Hubble tension. Also,
we’ll talk about how the Surprise behaves when we analyze distributions with weak
non-gaussianities, compared with an analytical solution for the Surprise for gaussian
distributions.
