mode_compute=6 : Is that theoretically defensive?

[earsmall]

Hi, I am conducting Bayesian DSGE estimation using dynare.
Recently, I thought that I can almost overcome the difficulty of finding posterior modes.
In most cases, mode-finding problem arose because of shocks.
However, in some models, I could not bring myself to find posterior modes successfully.

What makes me reluctant is that I do not know exact rationale behind mode_compute=6 and
I do not know whether results obtained by using mode_compute=6 can be written in my paper.
Are there any papers that provide reasoning for monte carlo optimization? (mode_compute=6)

[jpfeifer]

mode_compute=6 is simply a MCMC-based global optimizer, not dissimilar to a simulated annealing. There is no reason not to use it. What you should always do is check whether you found the mode. In particular, try whether other optimizers find higher likelihood values.