In the command “stoch_simul”, there is an option called “drop” which represents the number of points dropped at the beginning of simulation before computing summary statistics. I wonder that what the use of “drop” is. I guess it is used to diminish the effect of starting distribution (burn-in sample) as in Metropolis Algorithm in Bayesian Estimation. Is it right?
I have looked for how IRF is computed and I discover “drop” is involved in determining the IRF of variable based on second-order policy function. But why we need to discard the path of variable for “n” time step, where “n” is the value in “drop”?
[quote=“dennis0125hk”]
I have looked for how IRF is computed and I discover “drop” is involved in determining the IRF of variable based on second-order policy function. But why we need to discard the path of variable for “n” time step, where “n” is the value in “drop”?[/quote]
You will find some explanations in the User Guide by Tommaso (section 3.8.2) available in the documentation section of www.dynare.org