Overview

Info-metrics – Definition

The available information is usually too complex and insufficient to deliver a unique solution for most economic modeling and inference problems. Info-metrics is a framework for consistently dealing with that problem. It is a framework for modeling, reasoning, and drawing inferences under conditions of deep uncertainty and limited information. It provides a rational inference framework for dealing with mathematically underdetermined (or partially identified) problems. Within a constrained optimization setup, info-metrics, in conjunction with information theory, provides us with a way to sort and rank solutions and choose the one that satisfies our desired properties. It provides us with a different way of thinking about solving such problems, handling model ambiguity and potential misspecification, and a way to nest models in terms of the information and the decision criterion they use. It also provides new insights into economic modeling and solves inference problems that cannot be solved with conventional methods.

The following two articles by Professor Golan provide a nice background to potential participants. Golan, A., 2018, Info-metrics for modeling and inference, Minds & Machines, is a very easy read describing the basic idea simply (no equations). Golan, A., Harte, J., 2022, Information theory: A foundation for complexity science, the Proceedings of the National Academy of Sciences (PNAS), describes the idea and recent development and some case studies (examples).

Objectives of this Short Course

In this short course we will concentrate on the study and practice of info-metrics inference. Though similar problems arise across most disciplines, we will focus on the study of Information-Theoretic (IT) modeling and inference in general with a strong emphasis on problems in the social sciences.

We will emphasize the fundamental theory and the motivation for using the theory, as well as how to practice it. We will practice the theory via computational experiments and with real data.

In addition, we will examine some of the key classical econometric methods and the way each one of them fits within that paradigm. Depending on time, we will also discuss the benefits of combining classical and information-theoretic econometric modeling, and whenever possible, the frequentist interpretation that info-metrics brings.

This class will be of interest for participants who did not see this material before and for those who already studied part of this material and are interested in a deeper knowledge and understanding of info-metrics (and information-theoretic methods of inference) in order to use and apply it for solving real world problems.

Note on Software and Code

Those who are used to write their own computer codes, the computing can be done by using different software, such as Matlab, GAMS, Python, R, etc. For those who wish to use common statistical/econometric software, all methods we discuss in this class are available in the main software packages, such as STATA, SAS and NLOGIT (LIMDEP).

All basic codes will be provided on the webpage: http://info-metrics.org/. A complete GAMS license will be provided for each class participant.

Key Words: Info-Metrics, Entropy, Inference, Information, Information-Theoretic Methods, Insufficient Information, Misspecification, Model Ambiguity, Uncertainty, Constrained Optimization, Economic Modeling