1 Dispersion and deviance residuals

1 Dispersion and deviance residuals For the Poisson and Binomial models, for a GLM with tted values ^ = r( X ^) the quantity D +(Y;^ ) can be expressed as twice the di erence between two maximized log-likelihoods for Y i indep˘ P i: The rst model is the saturated model, i.e. where ^ ................
................