Question regarding Stochastic Value Function Iteration

Hi Everyone, I am trying to solve a stochastic neoclassical-growth model with value function iteration. For some reason which I do not under stand, I get the following error message

if true
  Subscript indices must either be real positive integers or logicals.

    Error in valfun_stoch (line 22)
    c = amat(j,1)*k0^alpha - k + (1-delta)*k0; % consumption

    Error in fminbnd (line 119)
    x= xf; fx = feval(funfcn,x,varargin{:});

    Error in ericsims_vfi_stoch (line 54)
    k1 = fminbnd(@valfun_stoch,kmin,kmax);
  end

what does this error message with the incorrent matrix indexing mean? I looked it up but it was not clear to me in this context here...

The main code is as follows

if true

clear all; 
close all; 

   tic 

   global v0 beta delta alpha kmat k0 a0 s kgrid amat prob
   % set parameters 

   alpha = 0.33; % capital's share 
   beta = 0.95; 
   delta = 0.1; % depreciation rate (annual) 
   s=2; 

   tol = 0.01; 
   maxits = 300; 
   dif = tol+1000; 
   its=0; 
   kgrid = 4; % grid points + 1 

   % first solve for steady state 
   kstar = (alpha/(1/beta - (1-delta)))^(1/(1-alpha)) % steady state k 
   cstar = kstar^(alpha) - delta*kstar; 
   istar = delta*kstar; 
   ystar = kstar^(alpha); 

   kmin = 0.25*kstar; % minimum 
   kmax = 1.75*kstar; % maximum 
   grid = (kmax-kmin)/kgrid; % grid 
   kmat = kmin:grid:kmax

   kmat = kmat';
   [N,n] = size(kmat);

   amat = [1; 1; 1]

   prob=[1/3 1/3 1/3;
   1/3 1/3 1/3;
   1/3 1/3 1/3;
   ]

   v0 = zeros(N,1);

   while dif>tol & its < maxits

    for j = 1:3

        for i = 1:N
        k0 = kmat(i,1);
        a0 = amat(j,1);
        k1 = fminbnd(@valfun_stoch,kmin,kmax);
        v1(i,j) = -valfun_stoch(k1);
        k11(i,j) = k1;
        end
    end
    ggg = v1 - v0;
    gg = vec(ggg);
    dif = norm(v1 - v0)
    v0 = v1;
    its = its+1
end

toc

end

and the valfun_stoch function is defined as

if true
  function val=valfun_stoch(k) 

% This program gets the value function for a neoclassical growth model with
% no uncertainty and CRRA utility

   global v0 beta delta alpha kmat amat prob k0 s kgrid 

   klo=max(sum(k>kmat),1); % identify the gridpoint that falls just below . .

   %..the choice fork
   khi=klo+1;

   % do the interpolation
   g = v0(klo,:) + (k-kmat(klo))*(v0(khi,:) - v0(klo,:))/(kmat(khi) - kmat(klo))
   gg = interp1(kmat,v0,k,'linear')

   c = amat(j,1)*k0^alpha - k + (1-delta)*k0; % consumption

   if c<0
     val = -8888888888888888-800*abs(c);
   else
     val = (1/(1-s))*(c^(1-s)-1) + beta*(gg*prob(j,:)')
   end

   val = -val; % make it negative since we're maximizing and code is to minimize.
     end

If anyone has a clue that would be awesome, thanks a lot Cheers

Answer 1

[Roger Stafford]

The problem would appear to be the variable 'j' which you have in the line

 c = amat(j,1)*k0^alpha - k + (1-delta)*k0; % consumption

The function 'valfun_stoch' does not know what its value is. You didn't make it a global variable and it doesn't carry over automatically from the for-loop where it is an index. The error message is a bit of an understatement, since the trouble is even more fundamental than not being "real positive integers or logicals". It is totally unknown to 'valfun_stoch'!

[derrick]

Hi Roger, thank you so much for the quick response!

I fixed it and I do not get this error anymore!

Now, I get another error which is obviously associated with 'fminbnd'.

The error looks like this

if true
  Error using horzcat
  CAT arguments dimensions are not consistent.

    Error using horzcat
    CAT arguments dimensions are not consistent.

end

[Stephanie Parsons]

Did you figure out why you were getting the error with 'fminbnd'? I'm having the same problem.