function y=delta(guess)
%This function takes in guesses for the unknown parameters n, ln(k0) and
%E/RTr, and returns the deviation between the model and the data.  We bring
%the data in through the "global" command:
global crpass cr0pass Tpass

Tr=298; %The reference temperature.
qr=0.1; %The flow rate (liters/min)
m=1; %The amount of catalyst (g)

n=guess(1); %The first parameter
k0=exp(guess(2)); %The second parameter
ERTr=guess(3); %The third parameter

miss=crpass-cr0pass+m/qr*crpass.^n*k0.*(Tr./Tpass).^n.*exp(-ERTr*Tr./Tpass);

%OK, the question is how to weight each of the data points.  As it is
%currently written, it tends to accentuate the weighting at higher
%temperatures where the conversion is the largest.  This is because cr is
%lowest, and it is multiplied by a large value to make it balance cr0.  It
%also places a stronger weight on the runs with higher intial
%concentrations.  On the other hand, for lower cr we should have more 
%accurate measurements (if the fractional error is fixed, for example).  
%Different weightings will yield different "solutions" for optimal parameters.

%A reasonable choice is to weight each of the runs with the inverse of the
%initial concentration.  This is essentially equivalent to assuming an error
%proportional to the concentration measured, and each data point should be 
%of O(1).  Thus:
miss=miss./cr0pass;

%and thus we get the objective function:
y=sum(miss.*miss);