[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## Re: non-linear regression

**From**: |
Michael Creel |

**Subject**: |
Re: non-linear regression |

**Date**: |
Thu, 26 Feb 2004 11:31:16 +0100 |

**User-agent**: |
KMail/1.5.4 |

On Wednesday 25 February 2004 17:04, Al Niessner wrote:
>* I am trying to fit a sine wave (A*sin(b*x+C) +D) to a set of data --*
>* least square. Is there an existing function in octave to help with this.*
>* I checked the documentation (online) and browsed the 'octave> help' but*
>* nothing jumped out at me.*
Other peoples comments about the fft, you can use BFGSMin from octave-forge to
fit this. The attached example FitSine.m generates the output FitSine.out,
also attached.
This regression model suffers from problems of local non-identification. For
example, when A is zero, neither b nor C have any effect on the sum of
squared errors. Or, when b is zero, A and C are not separately identified.
This can cause problems when attempting to fit. Multiple random start values
will get around this.
Michael

**
**`FitSine.m`

*Description:* Text Data

**
**`FitSine.out`

*Description:* Text document