Quasi-Likelihood Estimation

A Spring 2021 AMS 573 group presentation explaining quasi-likelihood estimation and its use with overdispersed Poisson and binomial generalized linear models.

Topics covered

• moving from fully specified likelihoods to mean–variance relationships;
• quasi-score equations and estimator properties;
• dispersion estimation and sandwich-style covariance adjustment;
• quasi-Poisson and quasi-binomial models; and
• worked examples using horseshoe crab count data and grouped teratology data.

The examples show how extra-binomial or extra-Poisson variation can inflate standard errors and confidence intervals relative to their conventional GLM counterparts.

Repository contents

• presentation.pdf — final slides
• analysis.Rmd — presentation source with analysis
• Crabs.dat — horseshoe-crab example data

Contributors and presentation roles

• Bridget Hyland — motivation and transition from likelihood to quasi-likelihood
• Kai Li — derivation
• Peng Fei Yao — estimator properties and variance
• Chad Gueli — quasi-likelihood for Poisson and binomial GLMs
• Vinny Yabor — worked examples