| Author | Message | ||
     Tony Davies (td) Moderator Username: td Post Number: 214 Registered: 1-2001 |
Hello Tatjana, Welcome to the group. Two things: one thought, one caution Have you tried plotting and looking at spectra after different processing? If you have a difference in particle size you do not want to be using processes designed to reduce particle size effects. Does straight PCA show up any separation? Caution. These are powerful techniques you must have sufficient nunmber of samples to cover variation between batches of Hy1 and Hy2. No need to worry too much about the wavenumbers if you are use to working with wavelengths they are just different measures of the same dimension. Good luck, Tony | ||
| Marion Cuny (marion) New member Username: marion Post Number: 3 Registered: 6-2009 |
Dear Tatjana, There are other pre-treatments that you can try out to get a better S/N ratio. I am thinking about scatter effect correction methods: MSC or EMSC as well as SNV. You can probably use PLS-DA (you can code your sample 0 for Hypromellose type 1 and 1 for Hypromellose type 2) and make a cross-validation to study the stability of the coefficients by jack-knifing. The stable coefficients correspond to the wavenumbers that are relevant to differentiate your 2 types of sample. I am working for CAMO so all that I described can be done in The Unscrambler but you may find your way with other package too. Have a nice day Marion | ||
| tatjana tomazin (tomazin) New member Username: tomazin Post Number: 3 Registered: 5-2006 |
Dear all, I�d appreciate your help regarding the following subject: - I would like to distinguish between two HYPROMELLOSE (difference in Apparent viscosity, Degree of substitution and Particle size distribution, however the results for all tree tests are very close). - I chose cluster analysis (PCA+SIMCA), the whole range, without pretreatments, First and second derivative,..etc, however I was not successful. - What is your suggestion for selecting wavenumbers/ range (in this case and in general). I have had experience only in NIR for identification, so I have never played with wavenumber (de)selection a lot. I know that in should include the right wavenumbers, to get the right answer /results. - We also tried to separate different batches inside of each type of HYPROMELLOSE, due to the bad /good dissolution results of finished dosage forms. I believe for this task also selection of crucial wavenumbers is very important. But of course, maybe it is not possible to separate selected batches at all. - Any idea and information regarding the wavelength selection and suggested literature, articles, to inmprove the calibration. Looking forward to your answer. Kind regards, Tatjana. |