Which covariance to use for the proposal density of the MCMC sampler


Could you advise on mcmc_jumping_covariance. In particular, when to use either of the following options: (i) hessian, (ii) prior_variance (iii) identity_matrix.

Thank you

If possible, you should use the Hessian at the mode. If that does not work, you can try one of the other two. But before switching, you may want to investigate the reason why the Hessian does not work.

Thanks. Could I also ask where estimated bounds for the mode (not mean) are located (if there is such a thing).

What do you mean with that?