Again help regarding IRFs

[syedhussain84]

I hope someone will help me out.

Suppose I am working with a simple RBC model (example 1 in the dynare examples folder). Suppose I look at the IRF of capital w.r.t a technology shock. My understanding is that the graph that we get is the mean of various impulse response computed by dynare. My question is, how can I access those various impulse responses, in particular I want to get variance of each point in the impulse response.

Please help!

Thank you

[SébastienVillemot]

Actually the impulse response function is the response of the model to one standard deviation shock at first period, and no shock in the future: so the graph you get is not a mean of various draws, it is the path corresponding to only one draw.

You can retrieve this path in the oo_.irfs structure.

My guess is that you want the mean of IRFs taking into account the possibility of shokcs in future periods, and averaging over them. Note that for first order approximation, this is the same than what Dynare does, because certainty equivalence holds. For higher order, this can indeed be different, and you have to compute it by hand (you can use the "simult_" function to help you compute simulations; you may also want to use pruning—see the corresponding option—in order to avoid explosive trajectories).