function lam = lam_est(y,e,W)

%LAM_EST estimates lamda in sp_lag by maximizing logL with direct 
%        search method.
%
%INPUT: (i)   y = data vector
%       (ii)  e = estimated error vector
%       (iii) W = weight matrix
%
%OUTPUTS: lam = estimated value

n = length(W) ;

u = ones(n,1) ;
  
lam = fmin('lam_lik',-1,1,[],y,e,W) ;