issue with lsqlin versus regress

Hello, I am bit confused why lsqlin and regress yield different results. Here is a worked example. The lsqlin does not solve at all maybe I am missing something in the specification

    load carsmall

    x1 = Weight;

    x2 = Horsepower;    % Contains NaN data

    y = MPG;

    X = [ones(size(x1)) x1 x2 x1.*x2];

    b = regress(y,X)    % Removes NaN data

    x0 = 0.1*ones(size(X,2),1);

    options = optimoptions('lsqlin','Algorithm','active-set');
    xx = lsqlin(X,y,[],[],[],[],[],[], x0, options)

b =

   60.7104
   -0.0102
   -0.1882
    0.0000

xx =

     0
     0
     0
     0

[Matt J]

Did you intend to attach carsmall.mat? Otherwise, we cannot repeat your test. You should probably also attach the results you got for b and xx, so we can know what you're seeing.

[hdg D]

Good point thanks Matt I am unable to attach it but its a matlab dataset so should be available

[Matt J]

What do you get from

 >>cond(X)

[hdg D]

you are right! doh! nearly singular matrix!