DYNARE USER GUIDE by Tommaso Mancini Griffoli, 2007

This page contents errata, comments, additional material and modifications required by chnages in Dynare since the writing of the User Guide

In order to get writing rights on this page, please, contact Michel Juillard

RBC_Monop_Det.mod, p. 32

• Simulating a deterministic model over 2100 periods is usually unnecessary, because all the action takes place in the periods following the shock. Depending on the inertia of the model (measured, for example, by the largest eigenvalue immediately below 1) and of the size of the shock, return to equilibrium may take a few dozen or a hunder periods, rarely more.
• In RBC_Monop_Det.mod, there is no display of the results. If you want to see some simulated series, adding the following code will plot the trajectory of c for the first 100 periods of the simulation

  dsample 1 100;
  rplot c;

-- MichelJuillard DateTime(2007-09-29T13:53:45Z)

RBC_Est.mod, p. 55

• 2000 Metropolis iterations are way too few and here one specifies mh_replic=2000 only so that you don't have to wait too long to see some results. In practice one has to use often 50000 iterations and sometimes 500'000 or 1'000'000 iterations!
• When comparing the posterior distributions, displayed at the end of the program, one observes that, for several parameters, they are very close to the prior. Either the priors are too tight or the corresponding parameters are ill identified. This could be further investigated by increasing the standard deviation of the priors.
• (p. 59) A clearly asymetrical (nonnormal) distribution isn't necessarily the sign that something went wrong. In DSGE models, some parameters can have very nonnormal posterior distributions.

-- MichelJuillard DateTime(2007-09-29T13:53:45Z)

fs2000ns.mod, p. 72-73

• No parameters have been assigned values when steady is called, therefore steady returns NAN values for all but the nonstationary variables that are arbitrarily set to 1. As we want to estimate the parameters, it is normal not to assign values for them before hand, but then steady shouldn't be used in this example

-- MichelJuillard DateTime(2007-09-29T13:53:45Z)

The ..._steadystate.m file, p. 24

• I added a test in steady and estimation to verify that the values returned by the ..._steadystate function is indeed a solution for the static model. If not, Dynare reports an error an stops. -- MichelJuillard DateTime(2007-10-01T13:17:58Z)