Page 1 of 1

Identification and data fitting

PostPosted: Thu Jun 08, 2017 7:11 am
by phuong.macro
Hallo Leute

I have a question relating to estimating DSGE model with Bayesian method

I wrote dynare file. At the the same time, I have data file. Before running estimating, what should I do? I mean do I need to conduct two following tests?
1. The test for identification problem
2. The test how actual data fit model. Normally, I saw literature that authors often report the actual data and predicted value (generated by Kalman Filter) on the same Graph to evaluate how the actual data fits the model

Re: Identification and data fitting

PostPosted: Fri Jun 09, 2017 5:11 am
by jpfeifer
1. It is always a good idea to check identification as it is almost costless in terms of effort and may help detecting pathological issues
2. Yes, it is a good idea to check the model fit. But I don't understand what you write about the Kalman filter. The Kalman filter will just decompose the observed historical observations in the contribution of different shocks. In-sample, the "fit" will be perfect. What you may compare is moments from simulated data to moments of the actual data.