least square fit matrix dimension errors

Hi,

I am trying to run a nonlinear regression

my code is the following:

gamma  = 3;                     
eps    = 1.5;                 
alpha  = 0.36;               
delta  = 0.06;                
beta   = 0.8930;
Abar   = 0.4477;
sigma  = 0.3726;
sigmaA = 0.05;              
rho    = 0.95;                
z      = 10;               

% initial Guess 

be1=[0.1690; 0.6491; 0.7317; 0.6477; 0.4509; 0.5470];
be2=[0.2963; 0.7447; 0.1890; 0.6868; 0.1835; 0.3685];
be3=[0.6256; 0.7802; 0.0811; 0.9294; 0.7757; 0.4868];

% Initialization

      T=100; 
      A=rand(T,1);
      eta=rand(T,1);

      kh=zeros(T,1)+1.4837;
      theta=zeros(T,1)+0.4026;
      psi1=zeros(T,1);
      psi2=zeros(T,1);
      psi3=zeros(T,1);
      v=zeros(T,1);
      Rk=zeros(T,1);
       Yi=zeros(T,1);
       Rap=zeros(T,1);
       fi1=zeros(T,1);
       fi2=zeros(T,1);
       fi3=zeros(T,1);
       theta_y=zeros(T,1);
       kh_y=zeros(T,1);
       v_y=zeros(T,1);

 % X - Data 

      Xx    = zeros(T,length(be1));

       for t=1:T

          Xx(t,:)=[1 log(kh(t)) log(A(t))...
           log(kh(t)^2) log(A(t)^2) ...
           log(kh(t))*log(A(t))]; 

%         Xx(t,:)=[1 log(kh(t)) log(A(t)) (eta(t))...
%          log(kh(t)^2) log(A(t)^2) log((eta(t)^2))...
%          log(kh(t))*log(A(t)) log(kh(t))*(eta(t)) log(A(t))*(eta(t))];

           psi1(t)=exp(Xx(t,:)*be1);      % data from the true function 

     end 

       % Y-data

       for t=1:T
        Rk(t)=1+alpha*kh(t).^(alpha-1)-delta;
        Yi(t)=(1-alpha)*kh(t).^alpha - alpha*kh(t).^(alpha-1)+eta(t);
        Rap(t)=Rk(t).^(1-gamma)+(1-gamma)*Rk(t).^(-gamma)*theta(t)*Yi(t)-0.5*gamma*(1-gamma)*(theta(t)*Yi(t)).^2;
        fi1(t)=(v(t).^(1-gamma))*(1+alpha*kh(t).^(alpha-1)-delta)*((1-alpha)*kh(t).^alpha-alpha*kh(t).^(alpha-1)+eta(t));
        end

     fun=@(be1,Xx)exp(Xx*be1); 
     [be_new] = nlinfit(psi1,fi1,fun,be1);

In the last line of code, I got always an error due to inconsistent matrix dimension multiplication. I understood the problem, however I did not manage to get it into a form that I could use the "nlinfit" built in function by MATLAB. Can someone help me on this ?

Many thanks

Answer 1

[Thorsten]

You forgot to define alpha, delta, gamma; if they are rand(T,1); you have to use alpha(t), delta(t), gamma(t).

[msh]

Those are just parameters that I forgot to put here. Sorry about this. I believe that I am not sure I understood the way nlinfit works.