# Bayesian estimation:Error message about The matrix must be positive definite

Dear al,
I get following error message ，when doing Bayesian estimation, and then the `identification`commend find `The steadystate file changed the values of parameter psi ` , how should I solve this problem?

``````Log data density [Laplace approximation] is NaN.

wrong use chol
The matrix must be positive definite.
``````

There are a few problems.

1. Your prior is too tight for the data at hand. In particular, the bounds prevent the estimation from matching the data, particularly the low shock volatilities.
2. Your steady state values for the observables are quite far away from the data means.
3. Use `prior_trunc=0`.