In addition to the covariance matrices obtained from this EM algorithm, we define more covariance matrices from the SFA results; specifically, the Q = 5 rank-1 matrices FqTLqTLqFq, with q = 1,…,Q. The rationale for adding these rank-1 covariance matrices is that the SFA factors may directly reflect effect patterns in the data, and if so these matrices will be a helpful addition. (If they are not helpful, they will receive little weight when estimating π).
