I have a question concerning the impulse response output from dynare++. I understand the dyn_irfm_eps_mean output as the average stochastic impulse response function. However I have no idea what dyn_irfm_eps_var is capturing. Is it the conditional variance IR?

Now I have a question: even at order=1, the variance is bigger than 0. According to the Dynare++ manual, the IRFs are to a 1 standard deviation shock. But as at order=1 due to linearity all IRFs must be identical given an identical shock size, the variance should be 0. What is done there?

In the linear case, it is easy to compute the expected IRF because the expectation of future shocks equals zero. But, if, as dynare++, you perform a finite number of simulations with different sets of future shocks, the empirical mean wonâ€™t be exactly equal to the theoretical IRF above and the variance will be different from zero.