Information Criteria Fail for Dynamical Systems: Sampling Rate and Dimension Dependence

TL;DR

This paper shows that traditional information criteria like AIC and BIC often fail for dynamical systems because of sample correlation issues, with failure depending on sampling rate and system dimension.

Contribution

It derives explicit formulas to predict when standard information criteria will fail in dynamical systems, guiding better sampling protocol design.

Findings

Information criteria depend on sampling rate and dimension in dynamical systems.

Explicit formulas predict failure regimes of AIC and BIC.

Sampling protocols can be optimized to avoid these failures.

Abstract

Information criteria such as Akaike's (AIC) and Bayes' (BIC) are widely used for model selection in physics and beyond, quantifying the tradeoff between model complexity and goodness-of-fit to enforce parsimony. However, their derivation assumes uncorrelated samples, an assumption systematically violated by dynamical systems data. Here, through analysis of simple but representative dynamical models -- exponential decay, harmonic oscillation, and chaos -- we demonstrate that model selection depends sensitively on sampling rate and system dimensionality. We derive explicit formulas predicting when standard information criteria fail that should be adaptable to many real-world scenarios, enabling experimentalists to design sampling protocols that avoid pathological regimes.