Maximum simulated log-likelihood values by quadrant. Real data, estimated optima, conventional Halton draws.

*This is the probability that the estimated Choleski factorization lies in this quadrant. The estimation is performed 52 times with different sets of starting values. **Only the difference between likelihoods is interesting, and the lowest estimated log-likelihood value (−8,388.20, from the estimation with 1,000 draws in Q2) is therefore subtracted from all the estimated values, in order to make the table easier to read. ***One of the four results in this quadrant differs from the others by 0.274. Maximum simulated log-likelihood values by quadrant. Real data, estimated optima, conventional Halton draws.