| Author | Message | ||
     Erik Skibsted |
I would like to know if anyone have a opinion on polar qualification system vs. other algorithms. I've made a simple matlab scribt that calculates the centre of gravities for the spectra and the plots looks very interresting compared to PCA/scoreplots when I want to discriminate between NIR spectra. But I am a novelty in NIR analysis and would like to hear the oppinion from all you with a lot more experience then I. Thank you. | ||
| Sheelagh Halsey (Sheelagh) |
The PQS system does do a nice job for qualitative applications, although it does tend to need a lot of input from the user. You generally have to optimise the spectral region to get a good discrimination or quality test. There is usually more than one way to solve a problem, PCs and wavelength distance are other algorithms I like. I would ususally look at several options to solve a problem. | ||
| Tony Davies (Td) |
You should get in-touch with Karoly Kaffka and his colleagues who developed the idea and have produced recently some software for utilising it. I have always thought it an interesting idea but have never done any work with it. I suppose I cannot see any reason why it should be an improvement on PCA but NIR spectroscopy is full of things which work quite well without much hard justification. The problem is that we still have too many ideas and no agreed method of classifying them! I think that one of the central problems with NIR information is that the way we collect the data as a spectrum is far from ideal. We really need an improved way of visualising the information. This might come out of the work being done by Don Dahm and others on diffuse reflection but I suspect we are still waiting for a new insight. | ||
| hlmark |
Tony makes a couple of good points, and I'd like to add that it's not unusual for there to be multiple solutions to any given problem - both in NIR and in other areas of life!!. So if anything, I consider that one of the strengths of NIR. Karoly is indeed the force behing the Polar Qualification System, which is recently being commercialized by a company called Metrika. I've always thought of the PQS as being an extension of the Mahalanobis Distance approach, modified so that you can use the entire spectrum, and don't have to select wavelengths, as the original application of MD required. Howard |