Skip to main content
. Author manuscript; available in PMC: 2010 Oct 26.
Published in final edited form as: Ann Appl Stat. 2010 Jun 1;4(2):916–942. doi: 10.1214/09-AOAS296

TABLE 3.

Computational procedure

  1. Initialize the parameter values:
    1. Choose an initial clustering. Two obvious choices are: (1) one cluster for all of the angle pair sequences, or (2) each angle pair sequence in a cluster by itself.
    2. For each initial cluster c of observed angle pair sequences, initialize the value of the common bivariate von Mises parameters µ, ν, Ω by sampling from the centering distribution H1(µ, ν)H2(Ω) of the DP prior.
      1. For the noninformative prior model, sample from each of m independent von Mises and Wishart distributions.
      2. For the DPM–HMM, obtain initial values for Ω from m independent Wishart distribution and µ, ν from the hidden Markov model.
  2. Obtain draws from the posterior distribution by repeating the following:
    1. Given the mean and precision values, update the clustering configuration using one scan of the Auxiliary Gibbs sampler of Neal (2000).
    2. Given the clustering configuration and mean values, update the precision matrix Ω for each sequence position in each cluster using the Wishart independence sampler described in Lennox et al. (2009b).
    3. If using the DPM–HMM, obtain a draw from the full conditional distribution of the state sequence s using the FB algorithm developed by Chib (1996) for each cluster.
    4. Given the clustering configuration, precision values, and (if applicable) state information, update the values of (μ, ν) for each sequence position in each cluster using the independence sampler given in Appendix B.