Algorithm 1 Algorithm 1 Blocked Gibbs Sampling for a Truncated Dirichlet process

1:

Input: N observations D = (x _i , t _i ).

2:

Initialize: c _i = random cluster assignments and parameters .

3:

Draw from the posterior of the joint distribution p(π, Φ*, c) by drawing from the conditionals.

4:

while NotCoverged do

5:

   Sample Φ* | π, c, D - This is carried out individually for each parameter in the model conditioned on the rest.

6:

   Sample c | Φ*, π, D - For i = 1,…, N, draw values , c _i = 1,…, M.

7:

   Sample π | Φ*, c, D - The mixture proportions are drawn based on the posterior P(π|α)P(c|π).

8:

end while