Dear members of the forum,
I guess the answer to my question is a straightforward yes or no. I have entered a set of non-linear equilibrium equations into my mod-file and competed IRFs using a first-order approximation. Now, I am wondering whether the same approximated conditions are used when I choose the option “linear_approximation” for a perfect foresight simulation?
Thanks in advance and kind regards,
Sebastian
Hi Seb,
I guess that should give the same IRFs after subtracting the steady state 
Ok, thanks a lot, Ben. In fact, I tried it out and came to the same result. So, yes: stochastic and deterministic simulations use the same linearisation when choosing the options accordingly.
Yes, because they also rely on the same code in the background.
Dear Johannes,
does that code actually compute numeric or analtic derivatives? Could you provide me with the name of the m.file?
Best,
Ben
The preprocessor does that. Dynare never uses numerical derivatives except for external functions. It always uses analytical ones.
Thanks for the confirmation, Johannes! Highly appreciated.