| Author |
 Topic  |
|
|
viswa21
Ireland
4 Posts |
 Posted - 11/07/2011 : 01:22:41 AM
|
Dear all,
I am trying to fit a curve shown in the attached file with a Gaussian-amp fitting function. I just want to fit the curve perfectly. How is this possible?? Any help would be appreciated.
PS: How to determine that Gaussian function is the one which fits perfectly??
Thanks.
please download origin file here
http://ifile.it/nh5oz2g/Bacteria spectra fit.opj
VISWANATH BAVIGADDA |
|
|
Penn
China
644 Posts |
Posted - 11/07/2011 : 10:16:43 PM
|
Hi,
Whether it fits perfectly or not, it depends on your data and the fitting model you choose. You can use the Residual Sum of Squares (RSS) to judge whether the fit is perfect. A small RSS indicates a tight fit of the model to the data.
However, I see that the RSS in your project is larger than 20000, which means that the Gaussian model is not so good to fit your data. Maybe you can change to other models to have a try again. I have changed the function to GCAS, and get a better result, but still not good enough, so maybe you need to define your own model according to your data.
Penn |
 |
|
|
viswa21
Ireland
4 Posts |
Posted - 11/09/2011 : 11:39:44 PM
|
Hi Penn,
I don't understand residual sum of squares, i do see regular residual, should i have to calculate the sum of squares manually and judge the fit???
Can you elaborate on what should be optimal conditions (parameters and their values) to fit curves with customized model?
I do see GCAS fits rather nicely, but residual variations gets worse than for Gaussian fit.
Thanks for lending me hand for learning some interesting things in origin.
Visu
VISWANATH BAVIGADDA |
|
|
|
Penn
China
644 Posts |
Posted - 11/10/2011 : 05:01:51 AM
|
Hi Visu,
The RSS is calculated by default. After fitting, you can find the RSS in the Statistics table of the report sheet. Of course, the RSS is not the only one method for judging the fit, you can use the method which you are familiar with.
If you define a model by yourself, maybe you need to pay more attention to two things: 1. the model should be suitable for your data. 2. provide the good initial value for each parameter. The initial values are very important for optimizing the fit. Sometimes the bad initial values will make the fitting not converged.
Penn |
|
|
|
viswa21
Ireland
4 Posts |
Posted - 11/11/2011 : 04:14:08 AM
|
Hi Penn,
I just again looked at fitting with GCAS function, but this time a little modification to the data, just normalized dependent variable between 0 to 1.
http://ifile.it/njdtr0g/Normalised bacteria spectra with GCAS.opj
Please have a look at the another project as well without normalization.
http://ifile.it/wjyu4kb/GCAS fit for bacteria spectra.opj
I do not know the theoretical background of the bacterial fluorescence to define own function to fit and also literature is little.
So would it be possible by changing some parameters value and try to minimize residual?? If yes plz let me know.
thanks
Visu
|
|
|
|
Penn
China
644 Posts |
Posted - 11/13/2011 : 9:20:31 PM
|
Hi Visu,
Theoretically, it will get a good fitting result if the good initial parameter values are provided. However, in your case, it seems that the result looks good now. So, if you still do not satisfy this result, I am afraid you had better define your own model for your data.
Penn |
|
|
| |
Topic |
|
|
|
Fitting a curve with data