Initial search volume in OSR

[Jingjing]

Dear all:
I am trying to use OSR to find optimal monetary policy parameters.
But I have got the following:

Error using cmaes (line 546)
Initial search volume (SIGMA) must be greater than zero

Error in Robot0924.driver (line 414)
[x_opt_hat, fhat, COUNTEVAL, STOPFLAG, OUT, BESTEVER] = cmaes('welfare_objective',x_star,H0,cmaesOptions,x_opt_name);

Error in dynare (line 281)
    evalin('base',[fname '.driver']);

[jpfeifer]

Your setting of H0 must be wrong.

[Jingjing]

Thanks for your reply!

I change the value of H0. And I use steadystate.m.

Dynare reports the following:

Does it imply the something wrong in steadystate.m?

[jpfeifer]

This looks like a version conflict. I would need to see the files.

[Jingjing]

Thanks for your reply, Prof. Pfeifer.

1. I run the mod file in Dynare 5.2, 5.0, 4.7-beta2, all three versions report following version:

csminwel

cmaes

2. The file runs well in Dynare 4.6.4, but the results of optimal monetary parameters seem weird which equal to upper bound.

3. I refer to other posts in the forum, I try to change the initial value, but the results also look weird in dynare 4.6.4.

4. Maybe there’s something wrong with my steady-state file, so I specify an initval block with the initial values for variables, however the optimal monetary parameters also equal to upper bound in dynare 4.6.4.

I send three files to you through P.M.

[jpfeifer]

Starting with Dynare 5, the call to resol has changed from

[oo_.dr,info,M_,options_,oo_] = resol(0,M_,options_,oo_); %get decision rules

to

[oo_.dr,info,M_,oo_] = resol(0,M_,options_,oo_); %get decision rules

[Jingjing]

Thanks for your reply, Prof Pfeifer.
I have changed the resol in welfare_objective.m.

But when use cmaes, the code could run, but it still show:

welfare_objective.m (1.2 KB)

2. The optimal monetary parameters still look weird:

phi_Pi=5; (5 is the upper bound)
phi_y=0.26; 

[jpfeifer]

1. As the message says, your bounds have width 5, but your H0 is just 0.02. Set it to 2 and the message goes away.
2. It’s quite common in the literature for optimal parameters to hit specified upper bounds.