rtist
in general, you can use multcomp package;
but for Duncan's, I don't think this package works;
nor do I believe Duncan't procedure is of any practical value.
If the procedure cannot control family-wise type I error rate, like Duncan's or SNK,
what's the point of doing any multiple comparison procedures? Just don't do anything
if power is your main goal! Otherwise, choose a procedure that does control family-wise
type I error rate.