[R] changing method of estimation in GLM

Ben Bolker bbolker at gmail.com
Sun Jan 2 16:03:07 CET 2011

ATANU <ata.sonu <at> gmail.com> writes:

> can anyone tell me how can i control the method of estimation (i.e. scoring
> method  or Newton raphson method) in glm and compute deviance function ?

  I don't think you can; you would have to write your own, although you
can take advantage of the framework of the current glm() [for interpreting
formulas, etc.]: see the "method" argument in ?glm.

  You might also find

install.packages("sos"); library("sos"); findFn("glm newton-raphson")

helpful; it finds a few special cases (logistic regression via Newton-Raphson)
in the logistf and glmmAK packages.

More information about the R-help mailing list