Goodness of fit

I’m trying to evaluate a new keynesian model, and make estimates with different variations around a “ground model”. Does anyone have any comments on how I should compare the different models and conclude which one is most likely the best by:

  1. Maximum Likelihood estimation
  2. Bayesian estimation

Is the smoothed shock graphs the only thig I can use for comparison?

Hi bk

Section 8.4 of User Guide has the answer to your question.

Try computing the posterior odds ratio between your “ground model” and the variation models. It should be easily done since Dynare outputs values of the marginal density of data conditional on the model.

Best,

8)