How to resolve the fminunc problem

Hey everybody, I am runing the post maximization of a financial friction model but each time I run it I have the following errors:

Error in ==> post_minus at 3
x = trans2(xtrans);

??? Error using ==> feval
Output argument "min_post" (and maybe others) not assigned during call to "F:\new
folders\67bis\Restricted Models (Table 3)\Fin Frictions Mod\Post
Maximization\post_minus.m>post_minus".

Error in ==> lineSearch>bracketingPhase at 103
          f_alpha =
          feval(funfcn{3},reshape(xInitial(:)+alpha*dir(:),sizes.xRows,sizes.xCols),varargin{:});

Error in ==> lineSearch at 48
[a,b,f_a,fPrime_a,f_b,fPrime_b,alpha,f_alpha,grad,exitflagBrckt,funcCountBrckt] = ...

Error in ==> fminusub at 208
    [alpha,f,grad,exitflagLnSrch,funcCountLnSrch] = ...

Error in ==> fminunc at 367
   [x,FVAL,GRAD,HESSIAN,EXITFLAG,OUTPUT] = fminusub(funfcn,x, ...

Error in ==> max_run_loop at 72
    [xtrans,fval,exitflag,output,grad,hessian]...

You can find the code following this link https://www.mathworks.com/matlabcentral/fileexchange/63625-the-financial-friction-post-maximization-code-of-chang-and-fernandez--2013-

When I run this code with the original data of the author it works and even with another set of variable, but using this data I didn't succeed to obtain the posterior mode and the inverse of the Hessian matrix to finish the estimation procedure (the original code can be found in the homepage of Andrés Fernandez).

Please hepl me

Thanks in advance

Houda

[Walter Roberson]

His page appears to be at https://sites.google.com/site/andresfernandezmartin8/research but I do not see links to code for this purpose there?

[Houda]

Dear Walter, yes right the website is"https://sites.google.com/site/andresfernandezmartin8/research" then you search for the title of his publication "On the Sources of Aggregate Fluctuations in Emerging Economies", the author mention the link of the code "Link to codes and data" click on https://sites.google.com/site/andresfernandezmartin8/Codes%20without%20MCMC.zip?attredirects=0 to download the code "codewithoutMCMC" then you move to restrected models=>financial friction model.

[Walter Roberson]

Could you clarify which routine you are calling, and with what parameters?

Answer 1

[Walter Roberson]

The code in post_minus includes the line

if logprior(1) == NaN

It is not possible to use "==" to compare to NaN . NaN == NaN is false. NaN has the strange property of being the only value for which x ~= x . The test should be

if isnan(logprior(1))

and likewise further down for elseif logdata(1) == NaN

Further down the code has

if ind == 1 % ind. is a flag signaling the inexistence of solution
   return;
end

It does this without having assigned a value to min_post. It should probably be

if ind == 1 % ind. is a flag signaling the inexistence of solution
   min_post = 0;
   return;
end

This is probably what you are encountering, that it is detecting that there is no feasible solution.

[Houda]

Dear Walter, it works you saved me, many thannnnnnnnnnnnnnnnks