[R] test of the effect of a polytomous explanatory variable (nominal scale) in glm with quasi

Eiiti Kasuya ekasuscb at kyushu-u.org
Wed Nov 17 17:03:37 CET 2010

When a polytomous variable (nominal scale) is used as an explanatory
variable in GLM (generalized linear models), the result object or
summary of glm function does not show the effect of this variable.  In
this case, anova function can be used. My question is on the appropriate
procedure(s) for testing the effect of the polytomous variable (nominal
scale) in glm with quasi family (not quasipoisson or quasibinomial, but
quasi).  Help of anova.glm says “for those with dispersion estimated by
moments (e.g., gaussian, quasibinomial and quasipoisson fits) the F test
is most appropriate”. In the case of quasi (not quasipoisson or
quasibinomial), should I use the following?

anova(result object, test = "F")

If there is another appropriate procedure of test in this situation,
please let me know.

Eiiti Kasuya

More information about the R-help mailing list