br.epsilon: positive convergence tolerance for the iteration described in brglm.fit.
br.maxit: integer giving the maximum number of iterations for the iteration in brglm.fit.
br.trace: logical indicating if output should be prooduced for each iteration.
br.consts: a (small) positive constant or a vector of such.
...: further arguments passed to or from other methods.
Details
If br.trace=TRUE then for each iteration the iteration number and the current value of the modified scores is cat'ed. If br.consts is specified then br.consts
is added to the original binomial counts and 2*br.consts. Then the model is fitted to the adjusted data to provide starting values for the iteration in brglm.fit. If br.consts = NULL
(default) then brglm.fit adjusts the responses and totals by "number of parameters"/"number of observations" and twice that, respectively.
Returns
A list with the arguments as components.
References
Kosmidis I. and Firth D. (2021). Jeffreys-prior penalty, finiteness and shrinkage in binomial-response generalized linear models. Biometrika, 108 , 71--82.
Kosmidis, I. (2007). Bias reduction in exponential family nonlinear models. PhD Thesis, Department of Statistics, University of Warwick.