mh_nblocks

[tandynare]

In the estimation command one can specify the mh_nblocks parameter. What does it really mean? In Metropolis-Hastings simulation, one can do it by separating the variables into several blocks/groups. Is this mh_nblocks parameter related to that, or is it something else?
Thanks.

[MichelJuillard]

mh_blocks indicate the number of parallel Metropolis chains that you want. It is only usefull to assess convergency

Best

Michel

[tandynare]

Dea Michel:
Thanks for your reply. That was helpful.
I have a few more questions regarding the use of estimation command.

1. From my limited understanding, the Metropolis-Hastings algorithm does not require information on a prior. Given so, is the specification of a prior in estimated_params important for simulating the posterior or for estimating the parameters?

2. Regarding the diagnostic figures produced by the estimation command, in the figure for the univariate diognostic, what are the two lines in the upper panel? I read the paper by Brooks and Gelman (1998), but still could not quite figure out what they were. What are m2 and m3 in the middle and lower panels. I suppose that the answer for the univariate figure is the same for the multivariate figure? Convergence occurs when all curves stabilize, I suppose?

Thanks in advance. Your help is always greatly appreciated.
Tan