[R] Question about SAMR

[Isabelle Rivals]

I have a question regarding the samr package.

For a 2 class unpaired problem, with sample 1 of size N1 and sample 2 
of size N2,  samr computes at most  (N1+N2)! permutations of the two 
samples (if the user-supplied parameter nperms allows it). However, 
there are only (N1+N2)/(N1!*N2!) DISTINCT permutations of the two 
samples, so it seems to me that only these distinct permutations should 
be taken into consideration. Of course, working with the (N1+N2)! 
permutations and working with the (N1+N2)/(N1!*N2!) distinct 
permutations will lead to the same correct result, but for a number of 
permutations smaller than (N1+N2)/(N1!*N2!), the result will generally 
be different.

I would be very  grateful if someone could provide me an explanation 
for the samr choice of the permutations.

Thanks for your help,
Isabelle


------------------------------------
Dr. Isabelle Rivals - Maître de Conférences
Équipe de Statistique Appliquée - ESPCI
10 rue Vauquelin - 75231 PARIS Cedex 05
Tél : (00 33 1) 40 79 45 45
Fax : (00 33 1) 40 79 44 20