Maximum Entropy, Time Series and Statistical Inference

TL;DR

This paper reviews the traditional Maximum Entropy Method, discusses criticisms, proposes a modified version, and demonstrates its application in statistical inference for time series analysis.

Contribution

It introduces a modified Maximum Entropy approach addressing prior criticisms and demonstrates its effectiveness in statistical inference tasks.

Findings

Modified method improves inference accuracy

Addresses criticisms of traditional maximum entropy

Demonstrates practical application in simple problem

Abstract

A brief discussion is given of the traditional version of the Maximum Entropy Method, including a review of some of the criticism that has been made in regard to its use in statistical inference. Motivated by these questions, a modified version of the method is proposed and applied to a simple problem, demonstrating its use in inference.