NonLinear Line Fitting - forced/constrained

Hi

I am really stuck on this. So I have two vectors

x=0:10; y=y+rand(1,11)*2;

Now I want to fit a variety of curves to this, specifically - linear,exponential,logarithmic,logistic and power.

My problem is what ever model i use I want to force the curve to go through {0,0} i.e constrain this point

I have been trying my hardest to figure out how to use fmincon but have had no luck. I have tried to use lsqlin (after linearising the models).

Can someone please give me a short example code using fmincon fitting a simple exponential model ae^{bx}? I think once I know how to fit one I can figure out all the rest :)

Thanks for any help

-----------------------------

I have figured out so far that to do this i need to use Aeq and beq. But I have been able to define a function in fmincon which can take two arguments

i.e

function f = myfun(p,x)
f = p(1).*exp(p(2).*x)

how do i pass my x vector into fmincon

Answer 1

[Fangjun Jiang]

Does this post http://www.mathworks.com/matlabcentral/answers/13255-least-squares-linear-fit-to-be-forced-to-pass-through-a-specified-point give you any clue?

[Sam]

thank you

For a simple linear regression i could just omit the intercept and thus force the line though zero. But if i want to implement a nonlinear model (one i can not linearise) then I need to use something more complex. From reading it seems that i should use fmincon