| Author | Message | ||
     Rseeth (Unregistered Guest) Unregistered guest |
I am tracking the concentration of a compound in a reaction with ATR immersion probe. I need to stop the reaction at roughly 50% consumption of the compound. We first ran the reaction three times, took samples at different times and got the samples tested by NMR to get exact converison rates. Then we built a PLS quant model and started using this model to predict the 50%converison point for reactions. Every once in a while our quant prediction is way off compared to NMR. The question is to explain why the quant is failing to predict. We take a solvent background before adding the compound into the reaction. I tried comparing solvent backgrounds from failed batches to good batches and could not find any trend. What are the chances for the ATR probe to drift from batch to batch? Is this a real possibility? | ||
| Suresh Kumar .BV (Unregistered Guest) Unregistered guest |
Hi Jon, As Solmon mentioned : Standard Normal Variate (SNV) is a row-oriented transformation which removes scatter effects from spectra by centering and scaling each individual spectrum. Each value in a row of data is transformed according to the formula: New value = (Old value � mean (Old row) ) / Stdev (Old row) You can choose MSC (Multiplicative Scatter Correction) to remove scatter effects. and also try with 1st Derivative | ||
| Solomon Abebe (Sol) New member Username: Sol Post Number: 4 Registered: 3-2006 |
Jon, If your software does not have SNV treatment option, you can do it using Excel. SNV is calculated as follows; Centre each spectrum data that means subtract the mean value form the original spectrum column data and then divide the mean centre data by the standard deviation of the mean centred data, take that new column spectrum as your pre-treated spectrum. However, you have to check other means of pre-treatment option as well. Because your spectral data may not need SNV treatment, it may need MSC etc. So, just find out which one is suitable for your data. | ||
| (Unregistered Guest) Unregistered guest |
Hi there, i've been reading a lot of papers where NIRS has been used and many people use standard normal variate and detrending to minimise the effects of sample size and maybe water as well. My OPUS software doesn't have these options and i was wondering if anyone new what the equivalent treatments would be with the following options i have: constant offset elimination, min-max normalisation, multiplicative scatter correction, vector normalisation, straight line subtraction, first derivative, scond derivative, 1st derivtive plus vector norm., 1st derivtive plus straight line subtraction....... I'm scanning whole leaves so i'm especially wanting to correct for scattering.I've had luck with various combinations but i was wondering if anyone new what would be similar to the aforementioned methods most comonly used in the literature. Any comments would be appreciated. jon |