 Weighted Least Squares by NIST

NIST/SEMATECH e-Handbook of Statistical Methods, section 4.4.3.2. The General WLS Criterion minimizes a weighted sum of squared residuals with fixed observations and weights. The centering identities in the article are derived explicitly from the intercept-only instance of this criterion, rather than asserted to be displayed by the source.

§4.1.4.3, Disadvantages of Weighted Least Squares: weights estimated from few replicates can adversely affect inference. §4.4.5.2, Direct Estimation of Weights and Estimating Weights with a Variance Function: replicate variances and smooth variance-function models are distinct weight-estimation approaches.