Comments on the solution to assignment 1. From the trace plots of lambda1 and lambda2, both Gibbs samplers appear to converge within a few tens of iterations. The autocorrelation plots for lambda2 show little difference between the two, and since the second Gibbs sampler takes about 6.5 times more compute time per iteration, the first Gibbs sampler seems preferable. On the other hand, we see that only the second Gibbs sampler has visited states where lambda1 is low (around 6) and lambda2 is high (around 7). So maybe the second Gibbs sampler has an advantage that is not apparent from the autocorrelation plot. As expected, one can see a negative correlation between lambda1 and lambda2, reflecting the fact that the counts can be explained as coming from either Professional Masters students or students from other departments.