Bias reduction for Binomial-response Generalized Linear ModelsOverview:
brglm provides a unified interface for estimation in binomial-response GLMs. Estimation can be performed using either an adjusted-score approach to bias reduction or maximum penalised likelihood, where the likelihood function is penalised by Jeffreys invariant prior, or simply maximum likelihood. The adjusted-score functions approach is the most appealing because the resultant estimator has zero bias to second order and smaller variance, and the bias-reduced estimates are always finite. Furthermore, brglm contains the tools for the construction of confidence intervals for the resultant estimates.