| Author | Message | ||
     Eric LALOUM |
Hi, In a specialized chemometric software BUCHI NIRCal, a specific indice is used to give the overall quality of a NIR calibration either quantitative or qualitative, called the "Q-value". For quantitative models, it is said to be calculated from : SEE, SEV, number of samples, number of terms... For qualitative models, it is calculated from the size and shape of clusters...It ranks from 0 to 1 and is used to compare several calibrations in an optimization algorithm. Does anyone have any experience with such indice ? Is it statistically funded ? What is the formula used ? Any articles regarding quality statistics for calibration ? Are there other quantitative quality indice to judge for the overall quality of a calibration (robustness and precision). Thanks, | ||
| maria verdelli |
We don"t have any experience about calibration model in Near analysis .Is really necessary in quantitative and qualitative analysis? Thanks. | ||
| bytecomm |
hello Maria Calibrations are derived from NIR absorbances in relation to the property values of a calibration set of data. This is cast in a background of molecular party chatter. All lighter molecules give up their ghost to NIR. You need calibrations to hear your ghost of interest. Happy Halloween | ||
| Luca Vicenzutto (Vice) |
HI! I currently use NIR from BUCHI and I also ask the same question to our "BUCHI man". He replies that Q value formula is covered by Industral Secret but is the result of a pondered average between SEP, SEE, Number of spectra, range covered. etc etc. In my opinion, I think that an High Q Value rappresents a good calibration. Regards | ||
| JEROLD CHICO (Jerold) |
Hi Vice! We are currently evaluating performance of BUCHI NIRLab N-200 and would also like to compare calibration generated against that of FOSS NIRS 6500. How is "Q-value" related to RSQ? Hope you could share more of your experiences with Buchi NIRS that will be helpful in our evaluation. Thanks and regards. | ||
| Luca Vicenzutto (Vice) |
Hi Jerold Some points: 1- We use BUCHI NIRFlex N-400, that is different than NIRlab. 2- What RSQ mean? Regards | ||
| Roland Winzen |
Hello all, if you're curious about the formula for Büchi's Q-Values you might be interested in the file CQValue.cls in the ..\Nircal 4\macros\ folder of your installation. (pure text files, basic-like code) In my experience the quantitative QValue is rather usable, as long as both calibration and validation set are _really_ representative for your data.. (yeah.. they always should be, I know...:-) Regards Roland Winzen http://www.nir-support.de | ||
| Luca Vicenzutto (Vice) |
Roland, thank you! I've found the document very interesting! Rgds | ||
| Luca Vicenzutto (Vice) |
Strange thing! I've update the software, but this file is disappeared! Could someone help me? | ||
| Tony Davies (Td) |
Vice, Please tell us what file you have lost! Best wishes, Tony | ||
| Luca Vicenzutto (Vice) |
Tony, I'm refering to CQvalue.cls that Roland indicated in his last reply! using Nircal latest version, this file is disappeared! Regards Luca |