Nonlinear Regression with ARMA Errors

Does Matlab have the capability of doing nonlinear regression where the errors follow an ARMA process? I need to determine significance of some of the parameters in a model and with correlated errors I have a highly inflated level of significance when the data is simulated under the Null hypothesis. As I read Seber and Wild (Nonlinear Regression) they indicate (page 285) that this is expected.

Answer 1

[Roger Wohlwend]

No, Matlab can only do linear regression with ARMA errors. Sometimes you can transform a non-linear equation into a linear one. If it is not possible in your case, you have to implement the regression yourself, using a non-linear optimizer. That sounds difficult, but it is not.