Find the Error - FZero Function. Constant loop.

[EDIT: Thu May 12 22:17:54 UTC 2011 Duplicate Removed, Reformat - MKF]

Okay so I can't seem to find this last error in my program. I just get a constant loop.

My first function file for Consumers reads:

function  F = Consumer2(x)        
%f1=N
%f2=n
%f3=k

global a R K W B w r

f1= (a*W)/(R*K + W*x(1) - x(3))+(a-1)/(1-x(1));

f2= B*((a*w)/(r*x(3) + w*x(2))+((a-1)/(1-x(2))));

f3= (-a/(R*K + W*x(1) - x(3))) + B*((a*r)/(r*x(3) + w*x(2)));

F= f1^2 + f2^2 + f3^2;

end

The second function file for firm profit reads:

function F = Firm2(x)        
%F1=N
%F2=n
%F3=k
global p r w K W

F1=.6*p*(K^.3)*((x(1))^-.4)-W;

F2=.6*p*((x(3))^.3)*((x(2))^-.4)-w;

F3=.3*p*((x(3))^-.7)*((x(2))^.6)-r;

F = F1^2 + F2^2 + F3^2;

end

Finally my executable is:

global a R K W p B r w

a=2/3;
R=.1409;
K=2;
W=.8;
p=1;
B=.95;
r=.5;
w=.7;
diff=1;
diff1=0;
diff2=0;
diff3=0;

while diff>.001

NCons=fminsearch('Consumer2',[.2,.1,.5])

NFirm=fminsearch('Firm2',[.2,.1,.5])

diff1=NCons(1)-NFirm(1);
W=W-.05*diff1;

diff2=NCons(2)-NFirm(2);
w=w-.05*diff2;

diff3=NCons(3)-NFirm(3);
r=r-.05*diff3;

diff=diff1^2+diff2^2+diff3^2;

end

When I run the executable i get an infinite loop. Any help?

[Sean de Wolski]

First: Don't call your variable diff since it's MATLAB's builtin difference function.

[Andrew Newell]

I believe that this code is an attempt to answer this question: http://www.mathworks.com/matlabcentral/answers/6442-optimization-problem-completely-lost-and-need-any-type-of-help. It would be nice if the two were merged.

Answer 1

[Andrew Newell]

The root of the problem is that you're trying to solve for NCons = NFirm by minimizing each of these functions at each step. The minimization is failing each time, leaving huge residuals. Here is a better way to solve it. Define your functions as

function F = Consumer2(x,a,R,K,B) %f1=N %f2=n %f3=k

f1= (a*x(4))/(R*K + x(4)*x(1) - x(3))+(a-1)/(1-x(1));

f2= B*((a*x(5))/(x(6)*x(3) + x(5)*x(2))+((a-1)/(1-x(2))));

f3= (-a/(R*K + x(4)*x(1) - x(3))) + B*((a*x(6))/(x(6)*x(3) + x(5)*x(2)));

F= f1^2 + f2^2 + f3^2;

and

function F = Firm2(x,K,p)

F1=.6*p*(K^.3)*((x(1))^-.4)-x(4);

F2=.6*p*((x(3))^.3)*((x(2))^-.4)-x(5);

F3=.3*p*((x(3))^-.7)*((x(2))^.6)-x(6);

F = F1^2 + F2^2 + F3^2;

(I have replaced W, w, r by x(4), x(5), x(6)). Then run:

a=2/3; R=.1409; K=2; W=.8; p=1; B=.95; r=.5; w=.7;
residual=1; residual1=0; residual2=0; residual3=0;

f = @(x) Consumer2(x,a,R,K,B)-Firm2(x,K,p);
xsol=fsolve(f,[.2,.1,.5, W, w, r]);

The line f = @(x) ... is what Matt means by points 2 and 3.

Answer 2

[Matt Tearle]

You're calling fminsearch with the same initial guess every time. So you get the same NCons and NFirm every time. So you get the same diff every time.

While I'm here...

1. As Sean says, don't use diff
2. Use anonymous function handles to embed parameters, rather than global
3. Use function handles, rather than strings for function names

[Jeremy]

Ty for the answer. I'm relatively new to the game here. This is my first time ever using matlab. It's for a class of mine. The professor used diff so I think I just sort of went with it. Not quite sure what the 2 and 3 mean... my apologies.