The precision of simulated series with simul_replic

[petros_varth]

Hi All,

I use stoch_simul to simulate series from my model, the economy fluctuates around its steady-state because of standard TFP shocks.

Particularly, the code is:

Code: Select all


stoch_simul(noprint,order=2,pruning,drop=0,periods=200,simul_replic=1000);



My question is the following:

Does the precision of the simulated series produced by the model increases as i increase the number of "simul_replic"?

If yes, in my model the weighted sum of period utility function (say for 200 hundred periods) is:

with simul_replic= 2000 ---> -767.6821

with simul_replic= 5000 ---> -755.7785

with simul_replic=10000 ---> -748.4416

with simul_replic= 20000 ---> -753.9464

with simul_replic= 30000 ---> -770.3927

Does the above look like converging?
Should i increase the number of simul_replic even more?
How i can use the same TFP shocks in each run?

Best,

P.

PS: The IRFs imply that the model makes sense.

[jpfeifer]

From the manual for simul_replic:

Number of series to simulate when empirical moments are requested (i.e. periods
> 0). Note that if this option is greater than 1, the additional series will not be used
for computing the empirical moments but will simply be saved in binary form to
the file ‘FILENAME_simul’. Default: 1.

Thus, for more precision, increase periods or use the get_simul_replicatons.m from my homepage to access the different replications of 200 periods and compute the moments yourself (https://sites.google.com/site/pfeiferecon/dynare)

[petros_varth]

Dear jpfeifer,

Many thanks for your reply. I will try this and post back.

P.