I am working with a model which doesn’t have a closed form solution for the steady state since it includes a probability density function with a variable used to compute this pdf and which has a value depending on the pdf value.
My current strategy to solve the model is to use a steady state file to compute the whole model, using a guess value for the annoying variable and running dynare to see if that guess is sufficiently closed from the true value to solve the model. But Dynare doesn’t like much pdf computation and I have to get a guess very close to the true value to be able to run the whole model.
I was wondering if there were a smarter way to do this (trying multiple guesses values for a steady state value), which wouldn’t take me about 15 minutes everytime I change a parameter in my model ?
Write an external Matlab file that runs Dynare once to initialize everything. Then use a solver directly on the steady state evaluation. Unfortunately, I would need to see the codes to provide more specific information.
Agreed with Johannes; check my answer in the other thread for further specifics: Problems in finding the steady state
I’ve tried using an external file linked to dynare, but never got it working. Although it’s ideal, I’ve found that a good steady state file that just gives “correct” values works if you initialize dynare at that point. If you figure out how to link an external steady state file, I’d like to learn too!
Thanks for the answers both of you, the link to the other thread is usefull too.
In my case, the easy way is to defined my guess parameter as global, and use an external file to run Dynare over a loop, using Dynare output to defined the next guess until it converges to the steady state.
It might be too early for me to invest lot of time on this issue since I’m still working on the specification and this issue is just slowing me down. I will face the problem with a nicer solution once I am sure the problem cannot be avoided.