Chunk #22 — 3.0 Summary of Small-Sample Covariance Estimators — 3.2 Mancl and DeRouen Estimator

Source

Modification of the Sandwich Estimator in Generalized Estimating Equations with Correlated Binary Outcomes in Rare Event and Small Sample Settings.

Embedded

yes

Text

where I is an nixni identity matrix, VN is the “naïve” or model-based variance estimator and Hi=D^iVND^iTV^i−1. Mancl and DeRouen justify this correction on the grounds that the true expected value is expressed as E[Cov(Yi)]≈(I−Hi)(Yi−μi)(Yi−μi)T(I−HiT) rather than E[Cov(Yi)]=(Yi−μi)(Yi−μi)T.