Counterfactual simulation, modes or means for parameters?

Estimation ML/Bayesian estimation
1

Dear all,

I want to do a counterfactual simulation of the model given smoothed shocks obtained after estimation. As I understand it, the smoothed shocks are obtained using the posterior mean estimates of parameters. So, my question is whether I should also use posterior means for parameters when doing the simulation? Or is it better to still use the modes?

Thank you for your help in advance!

2

If you run an MCMC, I would go for the posterior mean of the parameters and the smoothed shocks at this set of parameter values.

3

Thank you for your reply, Prof. Pfeifer!