Multivariate filter. Hessian is NPD

Hello, I am trying to replicate simple structural model from paper. However, estimation command returns a non positive definite Hessian in mode after the convergence of optimizator and then halts.
I also checked Blanchard-Kahn conditions in the model with prior means through identification command. It says that conditions are not satisfied. This is strange because I use prior means from original paper. In the internet there is also the code from the authors in Iris toolbox. Prior means in this code are the same as in the paper and in my mod file. This should be mentioned that in the original paper regularized likelihood was used.

I wonder if I need to switch to other mode optimization algorithm or there is something wrong in the way the model is defined in my mod file?
I am new to Dynare.
data.xlsx (20.5 KB)
filter_us.mod (2.9 KB)

Your data treatment seems to be wrong. The data is still trending and has a very different mean than the model variables.

Thank you for reply, proffesor!
I got it

Proffesor, sorry for disturbing you again,
I solved problem with trends in data, but I am still getting the same error.
data.xlsx (20.9 KB)
filter_us.mod (2.9 KB)

It works for me with
filter_us.mod (2.9 KB)
and the mode-file is
filter_us_mode.mat (20.9 KB)