Dear Johannes,
I and my coauthors have estimated the baseline DSGE model and its variants, and use Bayes factor do conduct Bayesian Model comparison between estimated baseline DSGE model and its variants, we find baseline DSGE model’s marginal likelihood is significant larger than those of its variants, however, we find model moments (simulated observable variables’ means and standard deviations) between estimated baseline model and its variants are different, although the differences are not very large? Does this sound awkward?
Our explanation is that all our data are in decimals but not percentualized, therefore although these differences do not look large, they maybe significantly different from each other, and we plan to conduct the following tests:
- Mean comparison t-test between baseline DSGE model’s moments of simulated observable variables and variant DSGE models’ moments of simulated observable variables.
- Standard deviation comparison F-test between baseline DSGE model’s moments of simulated observable variables and variant DSGE models’ moments of simulated observable variables.
Are we on the right track or not?
Thank you very much and look forward to hearing from you.
Best regards,
Jesse