Hi, everyone, I got wrong in estimation with the message “Error using chol. Matrix must be positive definite.” .

I have put identification command in the mod file, and it pass the identification check.

But still it comes up with something wrong in estimation .

Here are the mod file and data file , in which I have processed raw data under the guidance of Prof. Pfeifer’s “A guide to specifying Observation Equations for the Estimation of DSGE models”.

credit.zip (8.29 KB)

- Use a different mode-finder that does not get stuck. mode_compute=5 seemed to work well for your model
- Your prior for

```
phi_pi,1.5,0.001,30,GAMMA_PDF,1.5,10;
```

is problematic. It has a large asymptote at 0, but the lower bound for it should be 1. Please use a prior that gives more weight to the determinacy region.

Thanks a lot ,Prof. Pleifer!

It seems work well on mode_computer=5.

So here comes a question, do I have to revise and check again and again to make that estimation work under every mode_computer from 0-10?

Is it right that only the estimation has passed through the 10 mode_computers, can we make a conclusion that it is a rightly set-up model?

The goal is to find the global mode. It is often useful to test several mode-finders (but not all of them; several Newton-type algorithms if often redundant). You could try running mode_compute=5 first, load the mode-file and then run mode_compute=9. That is often sufficient.