MATLAB Answers

0

Error model with curve-fitting

Asked by Minu Pilvankar on 8 Oct 2019
Latest activity Commented on by Minu Pilvankar on 10 Oct 2019
Can I use one of the error models in Simbiology (constant/proportional/combined) while estimating parameters in MATLAB? I am using lsqnonlin for curvefitting and estimating the parameters. How can I incorporate a constant error function with it?

  0 Comments

Sign in to comment.

1 Answer

Answer by Arthur Goldsipe on 9 Oct 2019
 Accepted Answer

SimBiology's error models are only available when fitting data to SimBiology models using functions like sbiofit. If you are directly calling lsqnonlin, then you are probably doing "traditional" nonlinear least-squares regression, which is equivalent to a constant error model. It is possible to adapt lsqnonlin for other error models, but it requires a fair amount of work. If you can't use sbiofit, then I would recommend using fitnlm (or perhaps nlinfit), if you have a license for Statistics and Machine Learning Toolbox. Both of these functions support proportional and combined error models.

  3 Comments

Thank you Arthur! yes, I am trying to do it the traditional way. I thought of fitnlm or nlinfit. But they do not let me add constraints while estimating parameters. So, i was opting to go with lsqnonlin or lsqcurvefit
Since I'm a developer on the SimBiology team, my other recommendation would be to use SimBiology. :-)
If that's not possible, you can take a look here to see how sbiofit reformulates other error models into a form that can be used with lsqnonlin.
Thank you Arthur!

Sign in to comment.