Extended techniques for generalized linear models (GLMs), especially for binary responses, including parametric links and heteroscedastic latent variables.