What does an Inf eigenvalue mean, in terms of modeling? I mean, what happens that we have a Inf eigenvalue?
How we should solve it?
When you are solving a deterministic or stochastic linear system (as we normally do in macro models), for example,
Ax(t+1) = Bx(t) + Cz(t), and matrix
A is singular, you may have unstable or infinite eigenvalues. According to the Blanchard Khan conditions, this is not necessarily a problem if the number of unstable eigenvalues equals the number of jump variables in the model.
Thanks for your reply.
My problem is that I do not have the variance of many variables. I want to know how to solve it.
Not sure I understand you correctly. You mean you don’t have variance of many variables after running your model in dynare?
yes. exactly. and I guess it because of the inf eigenvalues.
Hello, you can try to run
model_diagnostics(M_,options_,oo_). I have had that issue before for many reasons, if I recall well, sometimes I have it when I miss an equation, hence, some equations show colinear relationship. I hope it can help.
As far as I know inf eigenvalues are not a proplem in and of themselves. Your missing moments could also come from the model featuring a unit root. Is that the case?
It was a problem in my code. I solved it. Thank you all!
For future reference, NaN moments do not come from infinite eigenvalues. They come from unit eigenvalues, i.e. unit roots.