| Author | Message | ||
     Judith Link (Linkj) |
I'm having a problem setting up a quantitative analysis of some products. I've done qualitative analysis and set up a project with constituents and values for all the constituents, I've tried to quantitate but cant seem to differentiate betweenthe components. There are three similar products: one with no added constituents, one with two added constituents and one with four added constituents. I have large numbers of samples---what % are optimally used for calibration? Any and all help will be greatly appreciated! The sofware used is Foss Vision. Thank you! Judith Link | ||
| Daniel J. Eustace (Eustacd) |
RE: Judith Link > >Our team published an article on what has worked for us with regard to NIR >analysis of complex mixtures, as Judith Link has asked. It is available on >line >http://pubs.acs.org/subscribe/journals/ci/31/i04/html/04f_chen.html > >The question she raises is more complicated and requires, as we have >learned, answers to several questions first. > >NO doubt others on this board have developed good approaches, as well. > >Hope this helps. >Dan | ||
| David Russell (Russell) |
Have you done the Quantitative Analysis Tutorial in Vision? That was necessary for me to get past "operator trouble". Otherwise, I've found that the NIRSystems regression techniques work reasonably well. So you may need an experienced colleague to look over your shoulder. | ||
| Sheelagh |
Where are you based? If you tell me your country and contact number, I can get a local support person to contact you. | ||
| Judith Link (Linkj) |
To Sheelagh, I'm based in the U.S. in Richmond Virginia----a little south of Washington,D.C. (linkj) | ||
| Sheelagh Halsey (Sheelagh) |
Hello Judith, since you are in the USA, the best route in to FOSS NIRSystems is through the web site, www.foss-nirsystems.com. There is technical support there if you give your instrument details, or there is a help desk number you can call. | ||
| Subodh |
Hi Judith, You may contact Dr. Jansen Christoph of Buchi -Switzerland and can give my reference to him. His e-mail is [email protected] and web site is www.ft-nir.com Subodh | ||
| Ronaldo Galv�o (Rgalvao) |
Dears I would like to know if they are companies working with quantitative analysis in the pharmaceutical industries and if it is possible according to the Pharmacopea transfer our methods like HPLC or UV to the NIR methods? | ||
| hlmark |
There is a two-part article published last year in J. Pharm. Biomed. Anal. that I wrote along with Gary and Emil, that addresses Ronaldos's question: J. Pharm Biomed. Anal., 28, p.251-260 (2002) J. Pharm Biomed. Anal., 29, p.159-171 (2002) I have sent Ronaldo e-copies of those articles. I hope they help Howard \o/ /_\ | ||
| Ronaldo Galv�o (Rgalvao) |
Dears We are working in a quantitative calibration and validation of Paracetamol in the Paracetamol tablets and we have found some difficulties in how to make the Accuracy in the method. I�ve found theories in the literature, but I don�t know how to calculate the Standard Error to prediction (SEP) and the Standard Error of cross validation (SECV)? What is the limit acceptable for the SEP? If I made a paired t-test it is enough? Thanks | ||
| Gary Ritchie |
Ronaldo, Here is a good reference for you: Mark, Howard, Workman, Jerry, Statistics in Spectroscopy, p. 299, 1981 | ||
| Sajeev Moorthiyedath (Sajeev77) |
Hi, I am trying to get information on SO2 detection and its limits in soy protein samples. How would I be able to do a spectral identification based on a set of previously run samples, working on WINISI software, even though I do not have a calibration set for the SO2. Thanks. Sajeev | ||
| Gabi Levin |
Hi Judith, One issue that has to be looked into is as follows: If you have a set of samples where all your constituents increase in value simultaneously it is very dangerous to create calibrations, because there is good possibility that you will pick on the change of one, and if the "slope" of increase of the constituents is similar for all of them, you will not be able to differentiate. To avoid this situation, you need sets where all constituents are present, but some increase, some decrease, some samples all constituents go up and some all are low, etc. Then, if you still get good regressions (assuming your constituents have a distinct spectrum, and that they are present in sufficient concentration) there is another thing you need to do - You have to look into the wavelength loading weights (feature easily available in Unscrambler) of the first few PC's and see if you identify spectral features that you can correlate to spectral features of the pure constituents. If the loading weights for all the regressions of all the constituents look very similar (sometimes mirror image) then there is a strong risk that your calibrations are not real. Mirror image loading weights occur most frequently when you have a "synthetic" set, where all variations in concentrations are linked to each other. If you find true differences in loading weights the next thing to do is to try and confuse the issue by running regressions on the same set, but choose few samples and "switch" values of constituents between tehm - if your regression is still as good, something is wrong, if the regression "senses" the disturbance then it is more likely to be reliable in real life. I would be happy to try a run on samples if you send them, and provide hopefully a better explanation. I hope this helps a little bit Gabi Levin | ||
| David W. Hopkins (Dhopkins) |
Hi Sajeev, Is that sulfite addition that you are interested in? I did not find any references to sulfite determination in the NIR bibliography of the CNIRS. It would certainly depend on levels of SO2 you are trying to measure, and the best bet would be to assemble a calibration set and try it. I learned a long time ago, never say "No, it can't be done." For starters, you could take scans of some samples before and after treatment and see if you see any spectral differences that seem correlated with SO2. What form is the SO2? Can you scan the "pure" SO2 and see any potentially useful bands? Then, what dilution is present in the soy protein, and would you expect to see it, compared with the instrument noise? These are some things you can do without assembling a calibration set. Best wishes, Dave | ||
| hlmark |
Judith - a little addendum to Gabi's advice: if you can calculate the correlation coefficients between the various constituents, that can tell you how close you are to being in trouble that way. This could be done using almost any statistical data analysis program package. The lower they are, the better off you are. A more sophisticated approach would be to calculate the multiple correlation coefficient among all the constituents, but that usually requires special software. If you contact me off the discussion group we can discuss this further, if you wish to pursue it. Howard \o/ /_\ | ||
| sajeev77 |
Hi, Thankyou Dave for that information. I am planning to pursue PhD program in NIR spectroscopy. I have a Masters degree in Chemical Engineering. Could someone suggest me some Universities that offer research in chemometrics and NIRS in the USA. I would prefer a chemical engineering program. Thankyou and regards, Sajeev Moorthiyedath | ||
| Marco Menchinelli |
Hi, I'm sorry to interrupt you with a stupid question, but could anybody please give me a hint in where to find a manual or something to get started with the use of Vision for NIR? I have to find a method to determine the % of actives in solid detergents but i don't really know where to start from.. Thank you! Marco Menchinelli | ||
| David W. Hopkins (Dhopkins) |
Marco, I think the learning curve for Vision is rather steep. But once you learn the steps, you should find it pretty "easy" to develop methods. There are old printed manuals, but now the information is in the program help file. If you go to Help on the main menu bar and look at the Index, you will see the topics that should get you started. Once you understand the structure of the program, you will find "hints" on what to do next come up to guide your work. The flow of the steps is not intuitive. It might be helpful for you to take the Foss training course. Is it given near you? Check with your representative or the Foss website. Best wishes, Dave | ||
| Marco Menchinelli |
Hi, Thank you David,I am usig the help file, but sometimes I find it a little difficult to find out what the correct steps are, even with the Hints.. I have found a sort of little "step-by-step" manual and I'll try to combine informations. Thank you for your help Marco | ||
| francesca lastrucci (Franci) |
Hi, I've a question about quantitative analysis: can I determine (with FT-NIR) the concentration of 0.1% in a water solution? Which is the range in water solution?( I'm curious) Thanks bye | ||
| hlmark |
Francesca - of course, a big question is WHAT do you want to determine the concentration of in water solution? The nature of the analyte is important, since that determines the strength of its absorbtion bands. In general, water is the most strongly absorbing material in the NIR region. It therefore follows that whatever is dissolved in the water will generally have less absorbance. That is not necessarily a showstopper, since absorbance bands of organic materials tend to be sharper than the bands of water, and therefore can be distinguished from it. However, 0.1% is definitely a marginal amount of material to determine by NIR, but might be doable if all other conditions are carefully controlled. One aspect of this is the possible use of the effect of the analyte on the water spectrum itself; in a fairly well-known paper, Tomas Hirschfeld determined the salinity of seawater by its effect of the water spectrum. That said, indirect effects of this sort are very tricky to utilize, and fraught with possiblities for errors to creep in. The water spectrum is affected by all sorts of extraneous things: temperature, pH, presence of ions, etc. All those things would have to be carefully controlled. You would also have to account for the effect of other materials in the sample itself. Bottom line: difficult but not necessarily impossible. Howard \o/ /_\ | ||
| David W. Hopkins (Dhopkins) |
Franci, I agree with Howard, and would like to add that the physiological level of glucose in the blood is about 50 to 300 mg/dL, and 100 mg/dL = 0.1%. There are still many companies spending many research dollars and hours pursuing this application. If this had been a problem of determining that level in well-defined clear solutions with no interfering compounds, the application would probably be done by now. If you want to determine some organic drugs in syrup formulations, this level is probably easy to achieve, depending on the absorbance of the material in solution. If you have a specific application in mind, the best thing to do is measure the material in pure solution at 0.1% and determine whether you obtain measureable bands. If so, then you can run a feasibility study to find out about how accurately you can measure the material. Hope this helps answer your curiousity. Regards, Dave |