Multiple linear regression with constraint

Hi,

I need some help with a code. I need to run a multiple linear regression for 4 variables (x1, x2, x3, x4) : y = a x1 + b x2 + c x3 + d x4 such that: b = 1 and c = ad

I know i need to use the lsqlin function but i don't really understand how it works.

Thanks for your help !

[Matt J]

b=1 is not really a constraint. It means its value is known. You should just exclude it from the list of unknowns and do the regression in terms of a, b, and d.

Answer 1

[John D'Errico]

No. You cannot use LSQLIN. Sorry, but not an option.

Why not? Because one of your constraints is nonlinear, i.e., c=a*d.

You will need to use FMINCON, minimizing a sum of squares of residuals. FMINCON can handle the nonlinear equality constraint.

[Aurélien]

Thanks a lot fot this quick reply! I am not familiar with mathematical formulation so I have difficulty to code with FMINCON. Is it possible to give me a hand to implement the code which answers to my question?