Transcending The Least Squares

TL;DR

The paper introduces the method of quasi-optimal weights as a versatile and asymptotically optimal alternative to least squares, effective for complex non-linear and non-Gaussian problems.

Contribution

It presents a comprehensive framework that generalizes least squares, offering improved optimality and flexibility for a wide range of statistical modeling scenarios.

Findings

Quasi-optimal weights achieve asymptotic optimality.

The method is transparent and adaptable to non-linear, non-Gaussian cases.

Provides a general alternative to traditional least squares methods.

Abstract

The method of quasi-optimal weights provides a comprehensive, asymptotically optimal, transparent and flexible alternative to the least squares method. The optimality holds for a general non-linear, non-gaussian case.