%least squares example 1
%this code creates "experimental data" with error and then fits a model to this data
global RU0 %global variables to use in functions
global RUeq

error = 100; %amount of experimental noise, to be multiplied by rand-0.5

RU0 = 50; %baseline RU value for an SPR experiment
RUeq = 1500; %equilibrium RU value for an SPR experiment
t = 0:50:600; %time vector, 0 to 600 s
    
%first, make the true solution (Rtrue) using the true parameters
%then make Rdata, our experimental data (Rtrue+noise)
koff = 0.005; %true parameter, in 1/s
for i=1:length(t)
    Rtrue(i) = RU0 + (RUeq - RU0)*exp(-koff.*t(i));
    Rdata(i) = Rtrue(i)+error*(rand-1/2);
end
    
%then, starting with the known initial conditions,
%and a guess for koff, 
%fit the model to the data and get a fitted value for koff
k = 0.1; %guess for koff
%fit for 1 parameter: koff
[fit_koff, Res, J] = nlinfit(t,Rdata,@SPR_function,k);

%and lastly, use your fitted koff to solve for R
for i=1:length(t)
    Rfitted(i) = RU0 + (RUeq - RU0)*exp(-fit_koff.*t(i));
end


figure(1);
plot(t,Rtrue,'b-');
hold on
plot(t,Rdata,'r*');
plot(t,Rfitted,'g-');
legend('true RU function','RU data','RU model from fitted koff');
xlabel('time');
ylabel('RU');
title('SPR dissociation data');