NIR Technology Australia               Suite 103, 56 Kitchener Pde

                                                                                                                                                Bankstown, NSW, 2200

                                                                                                                                                Ph: 02 9708 5068

                                                                                                                                                Fax: 02 9708 5537

 

Application Note 28: Evaluation of Chemical Lean Analysis

 

Introduction:

 

An existing Chemical Lean calibration model was used to evaluate the performance of the NIT-38 Meat Analyser against a second company’s meat samples analysed for CL using the Microwave method.

 

The first plot shows the relationship between the existing NIR calibration vs the Microwave method. The R2 of .892 is not too bad but the scatter is high and the error is approximately 2.3 units. Our experience is that the two should agree to approximately 1unit.

 

 

 

Plot of existing NIR Calibration for Chemical Lean vs the Microwave Method

 

 

The next two plots show the result of calibrating the NIT-38 Meat Analyser against the Microwave method. The first plot is the original calibration data. I have used the 5 scans saved for each sample. As such there is a considerable amount of scatter. This is to be expected because there is going to be a lot of variation in the sample, as well as sample packing variations. Note the SEC(Standard Error of Calibration) is approximately 1.4 units and the R2 = 0.97. This is very good.

 

The last plot shows the effect of applying the calibration against the average spectra. The SEC is now 0.75 and the R2 = 0.99.

 

Plot of NIR vs Microwave Chemical Lean Calibration Data

 

 

Plot of NIR vs Microwave Chemical Lean, Prediction Data

 

Of course there are only 22 samples and I would like to have many more. However it demonstrates that the Meat Analyser can be easily calibrated to the Microwave method and that the error between the two methods is less than 5% relative error.

 

I have prepared a new calibration file for you. Please download this file into the instrument and use it for the rest of the study.

 

Can I also ask that the operators repeat the analysis, ie, do 1 set of 5 scans and then repack the sample and repeat it. They need not clean the cell between the replicates. This will provide more data to calibrate with, and allow us to assess the reproducibility of the calibration.

 

Regards

 

 

 

Phil Clancy

MD