Deterministic simulation and homotopic approach


May I ask two questions about the computation of transition path inside Dynare:

  1. If I know the actual path, how can I give to Dynare as the initial guess for deterministic simulation?

  2. Regarding homotopic approach (used to compute transitional path rather than steady state), how can I change the tolerance, iteration maximum, steps, etc to improve the accuracy?

I am currently working on a perfect foresight model. The model has already been calibrated by a set of given data. Then I used those exogenous variables and parameters calibrated by the data to conduct deterministic simulation. In theory, the transitional path simulated is supposed to be the same as the actual data.

Dynare cannot solve it, hence switch to homotopic method and succeed finally. But unsatisfactory part is that although most simulated variables equal to actual data, the state variable is slight different and inaccurate.

I guess the problem could be that the shocks in my model are too large and frequent. Hence the initial guess of transition path may be too far away from the SS, which causes Newton method to break down. As such, I come out with above two questions/ways to improve the computation accuracy.

Please kindly correct me if my guess is wrong and provide some suggestions on the computation.

Thank you very much!

  1. Use the new syntax for instead of simul. First call perfect_foresight_setup. Then set oo_.endo_simul to the starting values. Then call perfect_foresight_solver.
  2. Unfortunately, those options are hardcoded. You would need to change the perfect_foresight_solver.m