Hi,
I would like to calculate the log-likelihood distance for my cluster result but I encounter some problem with the formula written in the Two Step Cluster document explaining the mathematical formula in Section 5.4.1 for this distance measure.
- As the distance for cluster i and cluster can be calculated by dist(i,j) = ξ(i) + ξ(i) - ξ(i,j), then does the distance to itself can be calculated by this formula which return non-zero value and it makes me feel unsecure with that result. Can I convert those diagonal element in pairwise log-likelihood distance matrix for all my cluster into zero in that case?
- There was 2 formulas in that document which are slightly different in calculating the log-likelihood distance. Which one is more correct to implement in this case? The Formula was stated in page 2 and page 11.
Thanks and looking forward on your reply.
------------------------------
Wen Kai Yeam
------------------------------