Author |
Message |
Mikhail (fgvbrt)
New member Username: fgvbrt
Post Number: 2 Registered: 2-2012
| Posted on Monday, February 13, 2012 - 6:05 am: | |
Howard, thanks for quick response. Sorry for stupid question. For detrend reason of difference in results is lack of accuracy. For derivative: y values differs by multiplicative constant in all diapason of wavelenghts. Topic can be closed. |
Howard Mark (hlmark)
Senior Member Username: hlmark
Post Number: 472 Registered: 9-2001
| Posted on Sunday, February 12, 2012 - 6:36 pm: | |
Mikhail - unless there's a bug in the programming, it doesn't matter which way you do it. The only thing that matters is that the wavelength intervals are uniform across the spectrum and that the actual wavelengths at which measurements are made are the same for all samples. Then you should get identical results using either formalism. If you want to be extra-sure, do it both ways for one set of data, and verify that you get identical results. If you don't get identical results, then you should contact your supplier (CAMO in the case of Unscrambler) and find out why not. Howard \o/ /_\ |
Howard Mark (hlmark)
Senior Member Username: hlmark
Post Number: 471 Registered: 9-2001
| Posted on Sunday, February 12, 2012 - 6:35 pm: | |
Mikhail - unless there's a bug in the programming, it doesn't matter which way you do it. The only thing that matters is that the wavelength intervals are uniform across the spectrum and that the actual wavelenths at which measurements are made are the same for all samples. Then you should get identical results using either formalism. If you want to be extra-sure, do it both ways for one set of data, and verify that you get identical results. If you don't get identical results, then you should contact your supplier (CAMO in the case of Unscrambler) and find out why not. Howard \o/ /_\ |
Mikhail (fgvbrt)
New member Username: fgvbrt
Post Number: 1 Registered: 2-2012
| Posted on Sunday, February 12, 2012 - 12:07 pm: | |
I have questions about correct NIR spectra processing. Now I use the program unscrambler for this purpose, most of all i am interested in detrend and Savitsky-Golay derivative. Help me please clarify the following parts of this methods: Detrend: the purpose is to shift all peaks to X- axis, for this purpose we build a regression I=f(w), where I - is our y-values (Log(1/R), w- wavelengths; and order of regression as a rule is taken to 2 (but this is not necessary). My question is what should we take as w, the actual wavelength or the number of wavelength, for example if i have wavelengths from 1000 to 2000 nm with 2 nm step, should I take in regression as independent variable the (1000; 1002; 1004...2000) or should i take (1;2;...501)? I ask this question, because in unscrambler by default preprocessing is made as independent variables equals to 1,2,...(numbers, not actual wavelengths values). This concerns detrend as well as derivatives. Because when we use Savitsky-Golay derivative, we also build a regression, and in this case what independent variable should be? Thanks in advance. |
|