The estimation mod is passed and has reported the bayes irf graph:

bayes1006.mod (6.2 KB)

data_20220914.mat (3.9 KB)

But the simulation mod is failed:There are 5 eigenvalue(s) larger than 1 in modulus

for 6 forward-looking variable(s),The rank condition ISN’T verified!

esti1006.mod (5.6 KB)

I choose the post. mean of the estimation mod, the simulation mod should pass and report the same irf graph as the the bayes irf graph in the estimation mod. But the simulation mod is failed(B-K condition), why?

Your model has a unit root from what I can see. That requires `diffuse_filter`

in the estimation. Leaving that option out allows this root to be counted as explosive. Thus, you have the impression that the model works. But the `check`

before `estimation`

already shows that there is an indeterminacy problem in your model.

Using diffuse_filter may have problem:

I have tried to use diffuse_filter in the estimation mod, but the irf graph of the stoch_simu mod(using post_mean) is not the same as the irf graph of the estimation mod. Especially ,the two graph is quite different.

The previous mod(using *diffuse_filter* in the estimation):

bayes0927.mod (6.7 KB)

reported bayes irf:(the first irf graph)

stoch_simu mod:

simu0927.mod (6.1 KB)

The two irf graph is quite different, which means stoch_simu cannot repeat the *diffuse_filter* in the estimation.

stoch_simu reported irf:(the first irf graph)

The two irf graph is quite different, which means stoch_simu cannot repeat the *diffuse_filter* in the estimation.

Most importantly, it seems you are working with very different shock sizes/shock standard deviations. The shape of the IRFs does not look too different.