If the ‘x’ parameters are real and ‘w’ is real, and ‘eq’ is complex, one option would be to create your data as separate vectors of the real part and the complex part, then fit the real and complex parts of your model to the real and complex parts of your data.
It would go something like this (assuming column vectors):
Data = [real(eq(:)) imag(eq(:))];
and the model would then be:
eq_mdl = @(x,w) [real(model(x,w)) imag(model(x,w))];
