A Deep-Dive into Generalized Least Squares Estimation

A detailed look at how to fit a robust GLS model on heteroskedastic, auto-correlated data sets

Generalized Least Squares (GLS) estimation is a generalization of the Ordinary Least Squares (OLS) estimation technique. GLS is especially suitable for fitting linear models on data sets that exhibit heteroskedasticity (i.e., non-constant variance) and/or auto-correlation. Real world data sets often exhibit these characteristics making GLS a very useful alternative to OLS estimation.

Motivation …

correlation deep-dives heteroskedasticity least linear regression squares