How to constrain the results calculated by "nlinfit" from complex number to real number?

Hello everybody,

Recently I used the mathematic function "nlinfit" to execute the fit. Unfortunately, the results show that the fitting coefficients are complex number, which does not have a physical meaning of my model.

I know there is one way to just use the absolute value to get real number after the fitting. However, it is just a trick! So how can I constrain it to real number before starting the fitiing?

[Torsten]

You will have to constrain the parameters E and B such that T_b-T^0 remains positive.

Best wishes

Torsten.

[John D'Errico]

Taking the absolute value does not provide a useful solution anyway, since the result is not a minimum sum of squares of errors.

As Torsten said, you need to prevent that term from going negative or zero. This is not possible to do using nlinfit of course, since nlinfit does not allow that class of constraint on the parameters.