matmin2

 [par,err,gof,stat] = matmin2(gof_func,inpar,freepar,func,y,...)

 Recode to use Matlab fmincon and fminunc functions
 CURRENTLY DOESN'T PROVIDE PARAMETER ERRORS

 gof_func is the goodness-of-fit rule to use - chisq, logl etc.
 inpar are the initial parameter values to start minimization from.
 freepar is a logical array indicating if the corresponding parameter
 par should be varied in the fit. If freepar=[] then all par are
 taken to be free.
 func is the name of an N parameter user supplied
 function to fit which must have the form [y]=func(par,x).
 y,... are the data required to calc gof. For logl
 minimum would be n,x - for chisq min would be y,e,x -
 but y,e,x1,x2 with [y]=func(par,x1,x2) etc is needed
 for some problems.

 par are the output parameter values and err the errors
 on those parameters.
 gof is the goodness-of-fit at values par
 stat is the status of the fit

 eg: [x,n]=hfill(randn(1,1000),100,-3,3)
     [p,pe]=matmin('chisq',[30,0,1],[],'gauss',n,sqrt(n),x)
     [p,pe]=matmin('logl',[30,0,1],[],'gauss',n,x)

0001 function [par,err,gof,stat] = matmin2(gof_func,inpar,freepar,func,y,varargin)
0002 % [par,err,gof,stat] = matmin2(gof_func,inpar,freepar,func,y,...)
0003 %
0004 % Recode to use Matlab fmincon and fminunc functions
0005 % CURRENTLY DOESN'T PROVIDE PARAMETER ERRORS
0006 %
0007 % gof_func is the goodness-of-fit rule to use - chisq, logl etc.
0008 % inpar are the initial parameter values to start minimization from.
0009 % freepar is a logical array indicating if the corresponding parameter
0010 % par should be varied in the fit. If freepar=[] then all par are
0011 % taken to be free.
0012 % func is the name of an N parameter user supplied
0013 % function to fit which must have the form [y]=func(par,x).
0014 % y,... are the data required to calc gof. For logl
0015 % minimum would be n,x - for chisq min would be y,e,x -
0016 % but y,e,x1,x2 with [y]=func(par,x1,x2) etc is needed
0017 % for some problems.
0018 %
0019 % par are the output parameter values and err the errors
0020 % on those parameters.
0021 % gof is the goodness-of-fit at values par
0022 % stat is the status of the fit
0023 %
0024 % eg: [x,n]=hfill(randn(1,1000),100,-3,3)
0025 %     [p,pe]=matmin('chisq',[30,0,1],[],'gauss',n,sqrt(n),x)
0026 %     [p,pe]=matmin('logl',[30,0,1],[],'gauss',n,x)
0027 
0028 if(isempty(freepar))
0029   freepar=ones(size(inpar));
0030 end
0031 
0032 if(size(inpar)~=size(freepar))
0033   error('freepar must be same size as inpar (or empty)')
0034 end
0035 
0036 opt=optimset('Display','off');
0037 
0038 if(sum(freepar)==length(freepar))
0039   [par,gof,stat] = fminunc(gof_func,inpar,opt,func,y,varargin{:});
0040 else
0041   lo=-inf*ones(size(inpar));
0042   up=+inf*ones(size(inpar));
0043   ind=freepar==0;
0044   lo(ind)=inpar(ind);
0045   up(ind)=inpar(ind);
0046   [par,gof,stat] = fmincon(gof_func,inpar,[],[],[],[],lo,up,[],opt,func,y,varargin{:});
0047 end
0048 
0049 err=zeros(size(par));
0050 
0051 return