Log data density [Laplace approximation] is NaN. Error using chol Matrix must be positive definite.

Estimation ML/Bayesian estimation
1

Dears,

I prepared dataset based on JPfeifer tutorial (A Guide to Specifying Observation Equations for the Estimation of DSGE Models). There are 4 observable variables:
1)real gdp per capita,
2)real government consumption per capita,
3)interest rate,
4)CPI (P_t/P_t-1).
I transformed mentioned variables as follows:
1)Y_L=LOG(Y) → extracted cyclical component (HP filter), which allows to remove trend
2)Y_G=LOG(G) → extracted cyclical component (HP filter), which allows to remove trend
3)CPI=CPI/100 → LOG(CPI) → extracted cyclical component (HP filter), which allows to remove trend
4)R(annualized, percentage points) → (1+R/100)^(1/4) → extracted cyclical component (HP filter), which allows to remove trend.

Afterwards I ran the code and obtain mentioned in the subject error. I also tried to apply solutions from the similar topics, but still it doesn’t work. Could you please help to solve this issue? Please find attached .mod file and dataset.
DSGE8.mod (1.7 KB)
mod_danee.xlsx (14.2 KB)

2

It seems there is a problem with the steady state of g. Why is it not mean 0? Also, you should not use the two-sided HP filter to remove trends for full information estimation of models with a state space representation.