Weighted-Average Least Squares: Model Averaging, Shrinkage, and the Normal Location Problem

Authors: Giuseppe De Luca and Jan  R. Magnus

Publisher: Chapman & Hall/CRC Press

Series: Monographs on Statistics and Applied Probability 

Planned date of publication: Spring 2027 

Table of contents [download] 

Short description: 

This monograph introduces a new model-averaging approach, called weighted-average least squares (WALS), which simultaneously accounts for the noise caused by model selection and the noise caused by estimation. WALS is efficient, performs well in practice, and is computationally fast. The theory of WALS is firmly based on the theory of the normal location model, which is discussed in detail and will be of separate interest. The book contains theory, applications, and software. 

Main points of interest:

·  A thorough discussion of the normal location problem, both from a Bayesian and a frequentist viewpoint.

·  General introduction to model averaging.

·  Detailed development of the WALS procedure.

·  Implementation of the WALS procedure in Stata, geared to both experienced and less experienced users.

·  Simulations, robustness checks, comparisons with other model-averaging estimators, and case studies.

Intended audience: 

The text is intended primarily for senior undergraduate and graduate students, and for academic researchers in econometrics, applied economics, statistics, and mathematics.