I am working with a medium-sized New Keynesian model for which Dynare produces nice impulse responses. The output shows the recursive laws of motion, but only the numerical values for the coefficients, not the general solution in terms of parameters. As I would like to look at the sensitivity of my results more closely and not just by producing more and more impulse responses, I am wondering whether there is any way to get the laws of motion in terms of parameters. Does anybody know how that can be done (without doing it by hand)?

I don’t think this is generally doable. Dynare is solving the model using the QZ decomposition, so you’d have to be able to do that analytically. Are there any paper that do what you propose you could point me to? (I think the answer is probably no… It’d be great to do what you want, but it’s not possible except in very special cases that you see in textbooks (e.g. the gali book)

Use the dynare command set_param_value. There are few examples of what you are looking for in the forum. Search the forum for the use of that command. It is of immense use.