Hi, There is something I am not quite getting here. I can follow through the answer to this, but am confused why we can't just use p(ij) = mu(ij)/lambda(i) to work out the transition matrix. Using that I get: P= 0 2/3 1/3 0 0 1 0 0 1 Then use (1/3, 1/3, 1/3)P to get the distribution at time 1. Where am I missing the point? Thanks...