The material says that there are k parameters used in estimation for treatment means(it makes it clear that there are k + 1 parameters but there's constraint
Σ[ n(i)* tau(i)] = 0 which restricts it to k).
While calculating SS(B) which is a variance the degrees of freedom get reduced to k - 1.How? To my mind reason is as we use another estimation μ(i) hat in the calculation of SS(B) the d.o.f get reduced by one more to k-1.
If anyone holds different opinion kindly reply with reasons justifying the same.
Last edited by a moderator: Sep 11, 2010