I want to simulate a model using a t-student (or Laplace) distribution for the shocks. I understand that by default Dynare uses gaussian shocks, as defined in the ‘simult.m’ file in the matlab folder.
What if I change the ‘simult.m’ file to incorporate a non-gaussian distribution (say Laplace) and then simulate the model under this specification? Is it fine or is there something wrong in it?
I get the point. I actually simulate the S&W (2007) with a second-order approximation.
Is it fair to use fat-tailed shocks with higher order moments or it’s just wrong?
You mean shocks with a zero third moment, but a higher fourth moment than the Gaussian distribution? That should be fine at second order.
Yes exactly. I do not allow for skewness but just for fat-tails.
I would like to try with a t-student distribution but I do not find a way to modify the line:
‘DynareResults.exo_simul(:,i_exo_var) = randn(nxs, DynareOptions.periods)’*chol_S;’
in the simult.m file. If you have any idea on how to do it, I would be really glad!