function[Bhat,Dhat]=fimlst(B,D,Omega,T,k1,k2)
% NOTE you must have the likeli.m file in the same folder for code to
% work!!!
% k1 is the number of uknowns in the B matrix 
% k2 is the number of unknowns in the D matrix which is assumed to be
% diagonal
% T is the sample size
% Omega is the matrix from first stage VAR
% B is the matrix where the unknowns must be specified in advance.  For
% example, suppose B=[1 -b12;-b21 1].  In the standard recursive var we
% assume that b12=0 thus b21 is the unknown.  We then specify the B matrix
% as B=@(x)[1 0;x(1) 1];  Recall D is assumed to be diagonal so that we
% specify it as D=@(x)[x(1) 0;0 x(2)].  For this example k1=1 k2=2;  For the
% order condition, we have to make sure that the number of unknowns in D
% which is n and number of unknowns in B does not exceed n(n+1)/2.  For our
% example, n=2 so k1+k2=3 so that we are just identified.  
n=size(Omega,2);
%out=fminsearch(@(x)likeli(B,D,Omega,T,k1,x),[ones(1,k1+k2)]);
%out=ga(@(x)likeli(B,D,Omega,T,k1,x),k1+k2);
out=patternsearch(@(x)likeli(B,D,Omega,T,k1,x),[ones(1,k1+k2)]);
Dhat=D(out(k1+1:k1+k2));
Bhat=B(out(1:k1));
save Dhat Dhat
save Bhat Bhat