In the theory and practice of econometrics the model, the methodand the data are all interdependent links in informationrecovery-estimation and inference. Seldom, however, are theeconomic and statistical models correctly specified, the datacomplete or capable of being replicated, the estimation rulesoptimal and the inferences free of distortion. Faced with theseproblems, Maximum Entropy Economeirics provides a new basis forlearning from economic and statistical models that may benon-regular in the sense that they are ill-posed or underdeterminedand the data are partial or incomplete. By extending the maximumentropy formalisms used in the physical sciences, the authorspresent a new set of generalized entropy techniques designed torecover information about economic systems. The authors compare thegeneralized entropy techniques with the performance of the relevanttraditional methods of information recovery and clearly demonstratetheories with applications including
* Pure inverse problems that include first order Markov processes,and input-output, multisectoral or SAM models to
* Inverse problems with noise that include statistical modelssubject to ill-conditioning, non-normal errors, heteroskedasticity,autocorrelation, censored, multinomial and simultaneous responsedata, as well as model selection and non-stationary and dynamiccontrol problems
Maximum Entropy Econometrics will be of interest to econometricianstrying to devise procedures for recovering information from partialor incomplete data, as well as quantitative economists in financeand business, statisticians, and students and applied researchersin econometrics, engineering and the physical sciences.